You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cocoon.apache.org by Marc Portier <mp...@outerthought.org> on 2004/04/20 15:17:32 UTC
Proposal: release management guide (was Re: [RT] Versions)
Carsten Ziegeler wrote:
> Marc Portier wrote:
>>in general I'm all for a more formal description of the
>>meaning of versions to us.
>>
>>something like: http://apr.apache.org/versioning.html but
>>probably more focussed towards Java classes, interfaces, xml
>>schema's and namespaces ?
>>
>>from there we could probably formulate some guidelines
>>towards associating code maintenance actions like
>>- bugfixing
>>- deprectation
>>- addition
>>- removal
>>
>>and how that is reflected in our cvs management towards
>>- tagging
>>- branching/merging
>>- new repository
>>
>>
>
> Yes, good idea!
>
> Do I understand you correctly that you are volunteering? Great :)
>
not really: I was hoping somebody was going to direct me to the URL of
the existing ones that could end my confusion :-)
in any case I'm not too shy to start this thread, and see where a joint
effort can bring us, I'm not the guy that wants everything carved in
stone, on the other hand it seems such a waste of time to discuss and
re-evaluate this kind of stuff on a case per case...
here goes:
I see some difference between what I would call 'extension' vs. 'usage'
compatibility
'usage' compatibility would be the guarantee that my xml files (elements
and namespace declarations) keep on being picked up by the machinery and
dealth with correctly (sitemap semantics, generator/transformer-picked
up elements, config file entries...)
it seems logic that this can grow, but should however remain backwards
compatible over the various minor releases of the same major release
parts of this usage we want to deprecate could be flagged through
run-time warnings to logger.warn() but remain supported (this would
indicate that an upcoming major release will no longer support this)
'extension' compatibility would be the guarantee that my own extensions
to what cocoon provides (own java classes that interface directly with
protected/package/public members in our distribution, and maybe also my
own javascript code that depends on cocoon provided library-functions)
maintaining these over various minor releases seems more then we need,
so keeping these stable within the scope of one minor release (ie across
the various patch-releases) should do. This would mean that parts of
that intreface can get deprecated but no additions/removals could happen.
I currently don't think we have a release scheme that supports one or
the other: i.e. reality seems pretty much like what Carsten is saying:
we just have 1,2,... 1004 (based on some gut feeling we seem to be
distributing those numbers over tripplets)
comments welcome,
-marc=
--
Marc Portier http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
Read my weblog at http://blogs.cocoondev.org/mpo/
mpo@outerthought.org mpo@apache.org
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Tim Larson wrote:
>
> On Thu, Apr 22, 2004 at 09:50:08PM +0200, Carsten Ziegeler wrote:
> > Tim Larson wrote:
> > > If it happens that 2.3 comes very soon after 2.2, then the
> > > deprecated code in effect was just deleted and not put through a
> > > normal deprecation cycle. Perhaps we need to also set a minimum
> > > length of time that deprecated features will live, to ensure that
> > > deprecation is a meaningful process and not just a formality.
> > >
> > Hmm, yes, this is a valid point. Looking back, it always took some
> > time between our releases, so I think we don't have to set a strict
> > rule for this. We should just decide on a case by case base if the
> > time inbetween releases has been long enough or if we
> should keep the
> > deprecated stuff in the next minor release and remove it one minor
> > release later.
>
> We will need to decide on a case-by-case basis, but we should
> also set a general guideline to help with the decisions.
Ok, we can add this when the guide is in CVS (which should be very
shortly)...
>
> > > Also, we should define what deprecation means, such as whether
> > > simple but severe security issues will receive updates,
> even though
> > > more ongoing or complex updates will not happen to
> deprecated code.
> > Can you please explain this a little bit? Do you mean, what will
> > happen if security issues arise in deprecated code?
>
> Yes, that is what I am asking. If we do not handle security
> issues in deprecated code, our users are not in a much better
> position than if we had just deleted the code without
> deprecation, since they can no longer safely continue to use
> the code in production. If we are going to do the effort of
> a deprecation cycle, we might as well make it mean something
> for our users.
>
Ah, yes, of course. If any severe security issue arises (where 'any' means
either in deprecated or non deprecated code), we will make
a patch release. Yes.
> > Yes, agreed. And we should really start updating libaries only if
> > there is really a good reason for it.
>
> For these compatible releases I agree, but I also appreciate
> Antonio's and others' efforts in keeping these libraries up
> to date. I wish there were an easy way to maintain the
> integration of both old stable and new up-to-date libraries
> against our code.
> I admit I do not know the best, most practical, balance for
> this issue.
>
Yes, in general I personally would always update to the latest
release, but for compatibility it's not always the best choice.
Now, I think Gump is doing this work for us, or? We use a
specific version and Gump builds (and tests?) Cocoon against
the latest version.
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Tim Larson <ti...@keow.org>.
On Thu, Apr 22, 2004 at 09:50:08PM +0200, Carsten Ziegeler wrote:
> Tim Larson wrote:
> > If it happens that 2.3 comes very soon after 2.2, then the
> > deprecated code in effect was just deleted and not put
> > through a normal deprecation cycle. Perhaps we need to also
> > set a minimum length of time that deprecated features will
> > live, to ensure that deprecation is a meaningful process and
> > not just a formality.
> >
> Hmm, yes, this is a valid point. Looking back, it always took
> some time between our releases, so I think we don't have to
> set a strict rule for this. We should just decide on a case
> by case base if the time inbetween releases has been long
> enough or if we should keep the deprecated stuff in the
> next minor release and remove it one minor release later.
We will need to decide on a case-by-case basis, but we should
also set a general guideline to help with the decisions.
> > Also, we should define what deprecation means, such as
> > whether simple but severe security issues will receive
> > updates, even though more ongoing or complex updates will not
> > happen to deprecated code.
> Can you please explain this a little bit? Do you mean, what
> will happen if security issues arise in deprecated code?
Yes, that is what I am asking. If we do not handle security
issues in deprecated code, our users are not in a much better
position than if we had just deleted the code without deprecation,
since they can no longer safely continue to use the code in
production. If we are going to do the effort of a deprecation
cycle, we might as well make it mean something for our users.
> Yes, agreed. And we should really start updating libaries only
> if there is really a good reason for it.
For these compatible releases I agree, but I also appreciate
Antonio's and others' efforts in keeping these libraries up to
date. I wish there were an easy way to maintain the integration
of both old stable and new up-to-date libraries against our code.
I admit I do not know the best, most practical, balance for this
issue.
--Tim Larson
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Tim Larson wrote:
>
> I am *so* happy to see this policy being worked on :) When I
> earlier proposed we write a document like this the interest
> seemed minimal, so I did not push very hard. Glad to see it
> is gathering steam now.
>
;)
> > Applications that write against a particular version will
> remain usage
> > compatible against later versions, until the major number changes.
> > Writing an application against a version means that this
> application
> > does not use any deprecated API of that version. Therefore minor
> > version changes are only usage compatible from one minor version to
> > the direct following one. For example 2.2 is usage
> compatible to 2.1.
> > But 2.3 is not necessary usage compatible to 2.1, although
> it is usage
> > compatible to 2.2. As long as you don't use deprecated API, your
> > application is usage compatible across all minor versions.
> >
> > Example:
> > - a feature is introduced in 2.0 and used by the application.
> > - it is deprecated in 2.2.
> > - it will be removed in 2.3.
>
> If it happens that 2.3 comes very soon after 2.2, then the
> deprecated code in effect was just deleted and not put
> through a normal deprecation cycle. Perhaps we need to also
> set a minimum length of time that deprecated features will
> live, to ensure that deprecation is a meaningful process and
> not just a formality.
>
Hmm, yes, this is a valid point. Looking back, it always took
some time between our releases, so I think we don't have to
set a strict rule for this. We should just decide on a case
by case base if the time inbetween releases has been long
enough or if we should keep the deprecated stuff in the
next minor release and remove it one minor release later.
> Also, we should define what deprecation means, such as
> whether simple but severe security issues will receive
> updates, even though more ongoing or complex updates will not
> happen to deprecated code.
Can you please explain this a little bit? Do you mean, what
will happen if security issues arise in deprecated code?
>
> > External Libraries
> > ------------------
> > Cocoon uses a set of external libraries (like for example Avalon,
> > Xalan or Xerces). Inbetween any release, even patch releases, the
> > versions of the external libraries might be updated to any version.
> >
> > Therefore if your application is written against a special
> API of an
> > external library it might be that this API of the external library
> > changes inbetween two Cocoon versions and therefore your
> application
> > does not work properly anymore (or even does not compile anymore).
> > Unfortunately, this issue is out of the scope of Cocoon.
>
> I echo Bruno's concern here. If we do an incompatible update
> of a library, and then cocoon starts to rely on the new
> features, then a user cannot just revert the library update
> to get a system as compatible as our versioning scheme would
> indicate we are shipping. To our users and developers it
> would feel like we broke our own versioning rules.
>
> Since *we* are the ones bundling the libraries I think this
> issue is in the scope of Cocoon. If a user were pulling in
> external libraries on their own, then that is what would be
> out of our scope.
>
Yes, agreed. And we should really start updating libaries only
if there is really a good reason for it.
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Tim Larson <ti...@keow.org>.
On Wed, Apr 21, 2004 at 09:33:14AM +0200, Carsten Ziegeler wrote:
> The Cocoon Versioning Manifest (CVM)
> ------------------------------------
> This document covers how the Cocoon project is versioned.
> Since Cocoon is a framework, it is very important to define
> a stable API for users and developers. However, we also need
> to move Cocoon forward, technologically. To balance these two
> needs, a strict policy of versioning is required, which users
> can rely upon to understand the limitations, restrictions,
> and the changes that can occur from one release to the next.
I am *so* happy to see this policy being worked on :)
When I earlier proposed we write a document like this the interest
seemed minimal, so I did not push very hard. Glad to see it is
gathering steam now.
> Applications that write against a particular version will remain
> usage compatible against later versions, until the major number
> changes.
> Writing an application against a version means that this application
> does not use any deprecated API of that version. Therefore minor
> version changes are only usage compatible from one minor version to
> the direct following one. For example 2.2 is usage compatible to 2.1.
> But 2.3 is not necessary usage compatible to 2.1, although it is
> usage compatible to 2.2. As long as you don't use deprecated API,
> your application is usage compatible across all minor versions.
>
> Example:
> - a feature is introduced in 2.0 and used by the application.
> - it is deprecated in 2.2.
> - it will be removed in 2.3.
If it happens that 2.3 comes very soon after 2.2, then the deprecated
code in effect was just deleted and not put through a normal deprecation
cycle. Perhaps we need to also set a minimum length of time that
deprecated features will live, to ensure that deprecation is a
meaningful process and not just a formality.
Also, we should define what deprecation means, such as whether simple
but severe security issues will receive updates, even though more
ongoing or complex updates will not happen to deprecated code.
> External Libraries
> ------------------
> Cocoon uses a set of external libraries (like for example Avalon,
> Xalan or Xerces). Inbetween any release, even patch releases,
> the versions of the external libraries might be updated to any version.
>
> Therefore if your application is written against a special API of an
> external library it might be that this API of the external library
> changes inbetween two Cocoon versions and therefore your application
> does not work properly anymore (or even does not compile anymore).
> Unfortunately, this issue is out of the scope of Cocoon.
I echo Bruno's concern here. If we do an incompatible update of a
library, and then cocoon starts to rely on the new features, then a user
cannot just revert the library update to get a system as compatible as
our versioning scheme would indicate we are shipping. To our users and
developers it would feel like we broke our own versioning rules.
Since *we* are the ones bundling the libraries I think this issue is in
the scope of Cocoon. If a user were pulling in external libraries on
their own, then that is what would be out of our scope.
--Tim Larson
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by David Crossley <cr...@apache.org>.
I think that we might mean "manifesto" (noun)
... Macquarie Dictionary:
"a public declaration ... of a body of persons taking
important action, making known intentions, objects,
motives, etc.; a proclamation".
Should it go into cocoon-2.1 CVS so that it travels with the
distribution, or just go into cocoon-site so that it ends
up at the top-level?
Carsten wrote:
> [ADD SOME REASONS WHY WE RECOMMENT RECOMPILATION ANYWAY]
See some notes at http://wiki.cocoondev.org/Wiki.jsp?page=Cocoon215TOC
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Joerg Heinicke wrote:
> > compatible to 2.2. As long as you don't use deprecated API, your
> > application is usage compatible across all minor versions.
>
> Somewhat inaccurate. IMO it must read "As long as you don't
> use deprecated API and the API you rely on does not get
> deprecated, your application is usage compatible across all
> minor versions."
>
Thanks, yes, good hint!
Carsten
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Joerg Heinicke wrote:
>
> > This high innovention has - at least in theory - the price of
> > maintaining
>
> innovation + invention = innovention ;-)
>
Ah, you spotted that one. Great :)
But it's not from me:
http://disneyworld.disney.go.com/wdw/parks/attractionDetail?id=InnoventionsE
astAttractionPage
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Joerg Heinicke <jo...@gmx.de>.
On 21.04.2004 09:33, Carsten Ziegeler wrote:
> Usage Compatibility
> -------------------
>
> 'Usage' compatibility guarantees that an application written
> by a Cocoon user is compatible. All files developed by a typical
> Cocoon user like xml files, sitemaps, stylesheets (elements and
> namespace declarations) keep on being picked up by the machinery
> and are dealt with correctly (sitemap semantics,
> generator/transformer-picked up elements, config file entries...).
> In fact this should cover everything (including flow script) but
> except own Java code.
>
> Applications that write against a particular version will remain
> usage compatible against later versions, until the major number
> changes.
> Writing an application against a version means that this application
> does not use any deprecated API of that version. Therefore minor
> version changes are only usage compatible from one minor version to
> the direct following one. For example 2.2 is usage compatible to 2.1.
> But 2.3 is not necessary usage compatible to 2.1, although it is
> usage compatible to 2.2. As long as you don't use deprecated API,
> your application is usage compatible across all minor versions.
Somewhat inaccurate. IMO it must read "As long as you don't use
deprecated API and the API you rely on does not get deprecated, your
application is usage compatible across all minor versions."
> Example:
> - a feature is introduced in 2.0 and used by the application.
> - it is deprecated in 2.2.
> - it will be removed in 2.3.
Otherwise this would oppose to this example.
> External Libraries
> ------------------
> Cocoon uses a set of external libraries (like for example Avalon,
> Xalan or Xerces). Inbetween any release, even patch releases,
> the versions of the external libraries might be updated to any version.
>
> Therefore if your application is written against a special API of an
> external library it might be that this API of the external library
> changes inbetween two Cocoon versions and therefore your application
> does not work properly anymore (or even does not compile anymore).
> Unfortunately, this issue is out of the scope of Cocoon.
Let's restrict this to updates of libraries without the need for
touching Cocoon code. This means the user must be able to revert the
update for his local cocoon installation.
> This high innovention has - at least in theory - the price of maintaining
innovation + invention = innovention ;-)
Joerg
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Marc Portier <mp...@outerthought.org>.
Reinhard Poetz wrote:
> Carsten Ziegeler wrote:
>
<snip type="large" />
>> Examples
>> --------
>> Here are some examples to demonstrate the compatibility:
>>
>> Original Version New Version Usage Compatible Extension
>> Compatible
>> 2.2.3 2.2.4 Yes Yes
>> 2.2.3 2.3.1 Yes No
>> 2.2.3 3.0.0 No No
>>
>> Note: while some of the cells say "no", it is possible that the
>> versions may be compatible, depending very precisely upon the
>> particular APIs used by the application.
>>
>>
>
> I think we will need *a lot* of testcases to guarentee the compatibility.
>
I agree, however on the sort-term and more pragmatic side of things: all
users out there are a very large test-cases-base, since we're up to
publishing this versioning contract to them we should at least promptly
react when people notice we've not lived up to it. (which should be:
re-introduce promised backwards compat and/or change the version number)
-marc=
--
Marc Portier http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
Read my weblog at http://blogs.cocoondev.org/mpo/
mpo@outerthought.org mpo@apache.org
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Reinhard Poetz <re...@apache.org>.
Carsten Ziegeler wrote:
>Marc Portier wrote:
>
>
>>I currently don't think we have a release scheme that
>>supports one or the other: i.e. reality seems pretty much
>>like what Carsten is saying:
>>we just have 1,2,... 1004 (based on some gut feeling we seem
>>to be distributing those numbers over tripplets)
>>
>>
>>
>I thought about this topic a little bit and I can only say we
>never followed any compatibility rules! For example we changed
>the sitemap in an incompatible way not only from 2.0.x to 2.1,
>but also from 2.1.3 to 2.1.4.
>And we removed deprecated API/changed interfaces/methods
>between patch releases.
>
>So following this "logic", we could just follow this road and do
>whatever we think is appropriate or establish some general and
>simple rules to make the live of the users easier.
>
>I have written a first draft of such a guide (well I copied some
>things from here and there and added my two cents). So, let's
>start another flame war and tell me what you think about this.
>
>
>(PS: I think this one of the longest mails I ever wrote)
>
>
;-)
>The Cocoon Versioning Manifest (CVM)
>------------------------------------
>This document covers how the Cocoon project is versioned.
>Since Cocoon is a framework, it is very important to define
>a stable API for users and developers. However, we also need
>to move Cocoon forward, technologically. To balance these two
>needs, a strict policy of versioning is required, which users
>can rely upon to understand the limitations, restrictions,
>and the changes that can occur from one release to the next.
>
>Basics
>------
>Versions are denoted using a standard triplet of integers:
>MAJOR.MINOR.PATCH. The basic intent is that MAJOR versions are
>incompatible, large-scale upgrades of the API. MINOR versions
>retain usage and extension compatibility with older minor versions,
>and changes in the PATCH level are perfectly compatible, forwards
>and backwards.
>
>
+1
>Following the main design principle of Cocoon, the pyramid
>of contracts, we distinguish between users and developers
>of Cocoon.
>A very rough distinction between them is that a user writes
>the application using Cocoon without coding Java. There is
>one exception to this rule: flow script - the java script
>is also written by the user.
>A developer codes Java and provides additional functionality
>for the user, so a developer extends Cocoon.
>
>Therefore we distinguish between usage compatibility and
>extension compatibility.
>
+1
>Both compatibility levels cover
>some kind of "source" compatibility. Cocoon does not provide
>binary compatibility. But as Cocoon is distributed as a
>source release that you have to compile anyway, it's
>saver to compile your own application code (if any) using
>the version of Cocoon that your application runs on.
>
>
+1
we should make a clear recommendation that recompiling is the better way
>Usage Compatibility
>-------------------
>
>'Usage' compatibility guarantees that an application written
>by a Cocoon user is compatible. All files developed by a typical
>Cocoon user like xml files, sitemaps, stylesheets (elements and
>namespace declarations) keep on being picked up by the machinery
>and are dealt with correctly (sitemap semantics,
>generator/transformer-picked up elements, config file entries...).
>In fact this should cover everything (including flow script) but
>except own Java code.
>
>Applications that write against a particular version will remain
>usage compatible against later versions, until the major number
>changes.
>Writing an application against a version means that this application
>does not use any deprecated API of that version. Therefore minor
>version changes are only usage compatible from one minor version to
>the direct following one. For example 2.2 is usage compatible to 2.1.
>But 2.3 is not necessary usage compatible to 2.1, although it is
>usage compatible to 2.2. As long as you don't use deprecated API,
>your application is usage compatible across all minor versions.
>
>Example:
>- a feature is introduced in 2.0 and used by the application.
>- it is deprecated in 2.2.
>- it will be removed in 2.3.
>
>However, if an application uses an API which has become available
>in a particular minor version, it (obviously) will no longer
>operate against previous minor versions.
>
>
+1
>Extension Compatibility
>-----------------------
>
>'extension' compatibility guarantees that own extensions to what
>Cocoon provides (own Java classes that interface directly with
>API in the Cocoon distribution) compile and operate.
>
>Applications that write against a particular version will remain
>extension compatible against later versions until the major or
>the minor number changes (Please note the difference to the usage
>compatibility). However, the Cocoon developers take care that even
>if the minor number changes, most of the own code still works and
>operates properly. Incompatible changes between minor versions are
>kept to a minimum. Frequent new releases of Cocoon ensure that
>developers have a smooth transition path.
>
>If an interface/class changes in an incompatible way between minor
>version changes, the Cocoon documentation will contain a detailed
>update guide that contains a solution for every incompatible change.
>So following this guide should make the update smoothly.
>
>
+1
>Deprecation and Exceptions
>--------------------------
>
>To continue the Cocoon development and to keep up with the innovations,
>parts of Cocoon might get deprecated; this includes parts of the user
>API and also parts of the developer API.
>
>If a part of the user API is deprecated, this will be flagged through
>run-time warnings that appear in the logs but remain supported. This
>indicates that an upcoming minor (or major) release will no longer
>support this.
>
>If a part of the developer API is deprecated it will be removed with
>the next major, minor or patch release.
>
>For developers there is one exception to this rule: private API. Cocoon
>has some internal classes and interfaces that are not meant to be used
>by a Cocoon developer (someone extending Cocoon). These pieces of
>Java code are clearly marked in the Javadocs and should not be used.
>They might change even between a patch version change in an incompatible
>way without providing a workaround!
>
>
+1
>External Libraries
>------------------
>Cocoon uses a set of external libraries (like for example Avalon,
>Xalan or Xerces). Inbetween any release, even patch releases,
>the versions of the external libraries might be updated to any version.
>
>Therefore if your application is written against a special API of an
>external library it might be that this API of the external library
>changes inbetween two Cocoon versions and therefore your application
>does not work properly anymore (or even does not compile anymore).
>Unfortunately, this issue is out of the scope of Cocoon.
>
>
I think we as Cocoon project can't avoid problems arising in this
context. So +1
>Examples
>--------
>Here are some examples to demonstrate the compatibility:
>
>Original Version New Version Usage Compatible Extension Compatible
>2.2.3 2.2.4 Yes Yes
>2.2.3 2.3.1 Yes No
>2.2.3 3.0.0 No No
>
>Note: while some of the cells say "no", it is possible that the versions
>may be compatible, depending very precisely upon the particular APIs
>used by the application.
>
>
I think we will need *a lot* of testcases to guarentee the compatibility.
>Versioning and Repositories
>---------------------------
>Cocoon is very innovative and new features are added very frequently. This
>results in new functionality that justify a minor version change on its own.
>In addition, to move Cocoon forward technologically, some features have
>to be removed/deprecated which of course results in a minor version change
>as well.
>
>So, Cocoon will potentially release several versions with minor version
>changes in a year and only a few patch releases. Of course, if a patch
>release is required, for example if a potential security issue is found
>etc., the patch release will be made available asap.
>
>This high innovention has - at least in theory - the price of maintaining
>several branches at once in order to be able to apply patches to older
>versions.
>To reduce this to a minimum, Cocoon uses one repository per major version.
>The HEAD of the repository always contains the latest version. If a new
>minor release is required, the version in the CVS is directly changed
>to this minor version (e.g. from 2.1.4 to 2.2). The current state is
>tagged and if the need for a 2.1.5 release arises, a branch is created.
>
>However, due to the compatibility rules explained above, there should only
>be a reason for a patch version if major problems occur (like security
>issues). In any other case, the Cocoon community expects the applications
>to migrate to the new minor version release which should be fairly simple.
>
>In addition Cocoon has a sandbox repository to test new features.
>
>
+1 again.
So we end in a cocoon-2 (current cocoon-2.1 repository) and a cocoon-3
repository (current cocoon-2.2 repository), don't we?
If we make move to SVN we can correct this naming issue. (It should also
be possible to change this in CVS too - only rename the directory names.
The only disadvantage is that we each CVS user is required to make a
fresh check out.)
--
Reinhard
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Unico Hommes wrote:
>
> For my understanding, what exactly do these terms mean (i.e.
> 'user API'
> and 'developer API') ? Are they paralel to the concepts of
> 'usage compatibility' and 'extension compatibility' or are
> they part of a separate classification?
Yes, they are parallel. Usage compatibility is the compatiblity
of the user api etc.
>
> >If a part of the user API is deprecated, this will be
> flagged through
> >run-time warnings that appear in the logs but remain supported. This
> >indicates that an upcoming minor (or major) release will no longer
> >support this.
> >
> >If a part of the developer API is deprecated it will be removed with
> >the next major, minor or patch release.
> >
> >
> >
>
> Don't you mean ".. removed with the next major or minor release" ?
>
No :) This is the tricky part. I think there are some very very very
very ... very rare cases where it makes sense to just remove some
deprecated stuff between two patch releases. So I don't want to block
this possibility. I will make this more clear in the next
version of this guide.
> >For developers there is one exception to this rule: private
> API. Cocoon
> >has some internal classes and interfaces that are not meant
> to be used
> >by a Cocoon developer (someone extending Cocoon). These
> pieces of Java
> >code are clearly marked in the Javadocs and should not be used.
> >They might change even between a patch version change in an
> >incompatible way without providing a workaround!
> >
> >
> >
>
> <snip type="more good stuff"/>
>
> As part of the effort of establishing versioning rules and
> consequently the separation of different areas of our code
> base (public, private,
> etc) we could review some of the related proposals that have
> been discussed in the past. Perhaps we could tackle these
> concerns as part of this effort. Especially I am thinking of
> the proposals towards modularisation of the code base into
> functional modules (API, SPI, etc.) and javadoc extensions
> for signalling private or public API membership.
>
> http://marc.theaimsgroup.com/?l=xml-cocoon-dev&m=105766229504616&w=2
> http://marc.theaimsgroup.com/?l=xml-cocoon-dev&m=106424060330963&w=2
>
Yepp, this has been discussed very often, so perhaps it's really
time to "do something".
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Unico Hommes <un...@hippo.nl>.
Carsten Ziegeler wrote:
<snip type="lots of good stuff"/>
>Deprecation and Exceptions
>--------------------------
>
>To continue the Cocoon development and to keep up with the innovations,
>parts of Cocoon might get deprecated; this includes parts of the user
>API and also parts of the developer API.
>
>
>
For my understanding, what exactly do these terms mean (i.e. 'user API'
and 'developer API') ? Are they paralel to the concepts of 'usage
compatibility' and 'extension compatibility' or are they part of a
separate classification?
>If a part of the user API is deprecated, this will be flagged through
>run-time warnings that appear in the logs but remain supported. This
>indicates that an upcoming minor (or major) release will no longer
>support this.
>
>If a part of the developer API is deprecated it will be removed with
>the next major, minor or patch release.
>
>
>
Don't you mean ".. removed with the next major or minor release" ?
>For developers there is one exception to this rule: private API. Cocoon
>has some internal classes and interfaces that are not meant to be used
>by a Cocoon developer (someone extending Cocoon). These pieces of
>Java code are clearly marked in the Javadocs and should not be used.
>They might change even between a patch version change in an incompatible
>way without providing a workaround!
>
>
>
<snip type="more good stuff"/>
As part of the effort of establishing versioning rules and consequently
the separation of different areas of our code base (public, private,
etc) we could review some of the related proposals that have been
discussed in the past. Perhaps we could tackle these concerns as part of
this effort. Especially I am thinking of the proposals towards
modularisation of the code base into functional modules (API, SPI, etc.)
and javadoc extensions for signalling private or public API membership.
http://marc.theaimsgroup.com/?l=xml-cocoon-dev&m=105766229504616&w=2
http://marc.theaimsgroup.com/?l=xml-cocoon-dev&m=106424060330963&w=2
--
Unico
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Reinhard Poetz <re...@apache.org>.
Upayavira wrote:
> Marc Portier wrote:
>
>>
>>
>> Upayavira wrote:
>>
>>> Carsten,
>>>
>>> Reads very well, I say. All we need to do is decide if that is how
>>> we actually want to (and can) work!
>>>
>>> One addition is to mention our policy on blocks and block status:
>>>
>>> Blocks and Block Stability
>>> --------------------------
>>> Cocoon currently allows optional functionality to be included or
>>> excluded using a simple system called blocks, in which the
>>> functionality is included or excluded at compile time.
>>>
>>> [NB. This is a precursor to a more complete block system which is
>>> currently under development.]
>>>
>>> A block can have one of three statuses: unstable, stable or
>>> deprecated. An unstable block has an API that can change without
>>> notice. A stable block is subject to the same versioning process
>>> as described in this document. Similarly, when the entire
>>> functionality of a block is deprecated, it will be handled in
>>> the same way as any other deprecated code within Cocoon.
>>>
>>>
>>> =========================
>>>
>>> Does this seem reasonable?
>>>
>>
>> very,
>>
>> I would even suggest (already) that blocks in state 'stable' would
>> maintain an own release numbering scheme which complies to these rules.
>
>
> I thought about this. But, given that they will be released as a part
> of Cocoon, I cannot see how that would work. Unless they are released
> as separate packages that happen independently, they'll need to follow
> Cocoon's versioning system, IMO.
>
> Regards, Upayavira
>
>
I think so too and IIRC this was also the conclusion of a discussion
about the lifecylce of _real_ blocks.
--
Reinhard
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Upayavira <uv...@upaya.co.uk>.
Marc Portier wrote:
>
>
> Upayavira wrote:
>
>> Carsten,
>>
>> Reads very well, I say. All we need to do is decide if that is how we
>> actually want to (and can) work!
>>
>> One addition is to mention our policy on blocks and block status:
>>
>> Blocks and Block Stability
>> --------------------------
>> Cocoon currently allows optional functionality to be included or
>> excluded using a simple system called blocks, in which the
>> functionality is included or excluded at compile time.
>>
>> [NB. This is a precursor to a more complete block system which is
>> currently under development.]
>>
>> A block can have one of three statuses: unstable, stable or
>> deprecated. An unstable block has an API that can change without
>> notice. A stable block is subject to the same versioning process
>> as described in this document. Similarly, when the entire
>> functionality of a block is deprecated, it will be handled in
>> the same way as any other deprecated code within Cocoon.
>>
>>
>> =========================
>>
>> Does this seem reasonable?
>>
>
> very,
>
> I would even suggest (already) that blocks in state 'stable' would
> maintain an own release numbering scheme which complies to these rules.
I thought about this. But, given that they will be released as a part of
Cocoon, I cannot see how that would work. Unless they are released as
separate packages that happen independently, they'll need to follow
Cocoon's versioning system, IMO.
Regards, Upayavira
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Marc Portier <mp...@outerthought.org>.
Upayavira wrote:
> Carsten,
>
> Reads very well, I say. All we need to do is decide if that is how we
> actually want to (and can) work!
>
> One addition is to mention our policy on blocks and block status:
>
> Blocks and Block Stability
> --------------------------
> Cocoon currently allows optional functionality to be included or
> excluded using a simple system called blocks, in which the functionality
> is included or excluded at compile time.
>
> [NB. This is a precursor to a more complete block system which is
> currently under development.]
>
> A block can have one of three statuses: unstable, stable or deprecated.
> An unstable block has an API that can change without
> notice. A stable block is subject to the same versioning process
> as described in this document. Similarly, when the entire functionality
> of a block is deprecated, it will be handled in
> the same way as any other deprecated code within Cocoon.
>
>
> =========================
>
> Does this seem reasonable?
>
very,
I would even suggest (already) that blocks in state 'stable' would
maintain an own release numbering scheme which complies to these rules.
-marc=
--
Marc Portier http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
Read my weblog at http://blogs.cocoondev.org/mpo/
mpo@outerthought.org mpo@apache.org
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Upayavira wrote:
>
>
> Reads very well, I say. All we need to do is decide if that
> is how we actually want to (and can) work!
>
Great! I think we could wait one more day and see if there is
some interesting feedback and then we can vote on it.
> One addition is to mention our policy on blocks and block status:
>
> Blocks and Block Stability
> --------------------------
> Cocoon currently allows optional functionality to be included
> or excluded using a simple system called blocks, in which the
> functionality is included or excluded at compile time.
>
> [NB. This is a precursor to a more complete block system
> which is currently under development.]
>
> A block can have one of three statuses: unstable, stable or
> deprecated. An unstable block has an API that can change
> without notice. A stable block is subject to the same
> versioning process as described in this document. Similarly,
> when the entire functionality of a block is deprecated, it
> will be handled in the same way as any other deprecated code
> within Cocoon.
>
>
> =========================
>
> Does this seem reasonable?
>
Yes, absolutely. Great!
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Upayavira <uv...@upaya.co.uk>.
Carsten,
Reads very well, I say. All we need to do is decide if that is how we
actually want to (and can) work!
One addition is to mention our policy on blocks and block status:
Blocks and Block Stability
--------------------------
Cocoon currently allows optional functionality to be included or
excluded using a simple system called blocks, in which the
functionality is included or excluded at compile time.
[NB. This is a precursor to a more complete block system which is
currently under development.]
A block can have one of three statuses: unstable, stable or
deprecated. An unstable block has an API that can change without
notice. A stable block is subject to the same versioning process
as described in this document. Similarly, when the entire
functionality of a block is deprecated, it will be handled in
the same way as any other deprecated code within Cocoon.
=========================
Does this seem reasonable?
Upayavira
Carsten Ziegeler wrote:
>Marc Portier wrote:
>
>
>>I currently don't think we have a release scheme that
>>supports one or the other: i.e. reality seems pretty much
>>like what Carsten is saying:
>>we just have 1,2,... 1004 (based on some gut feeling we seem
>>to be distributing those numbers over tripplets)
>>
>>
>>
>I thought about this topic a little bit and I can only say we
>never followed any compatibility rules! For example we changed
>the sitemap in an incompatible way not only from 2.0.x to 2.1,
>but also from 2.1.3 to 2.1.4.
>And we removed deprecated API/changed interfaces/methods
>between patch releases.
>
>So following this "logic", we could just follow this road and do
>whatever we think is appropriate or establish some general and
>simple rules to make the live of the users easier.
>
>I have written a first draft of such a guide (well I copied some
>things from here and there and added my two cents). So, let's
>start another flame war and tell me what you think about this.
>
>
>(PS: I think this one of the longest mails I ever wrote)
>
>
>The Cocoon Versioning Manifest (CVM)
>------------------------------------
>This document covers how the Cocoon project is versioned.
>Since Cocoon is a framework, it is very important to define
>a stable API for users and developers. However, we also need
>to move Cocoon forward, technologically. To balance these two
>needs, a strict policy of versioning is required, which users
>can rely upon to understand the limitations, restrictions,
>and the changes that can occur from one release to the next.
>
>Basics
>------
>Versions are denoted using a standard triplet of integers:
>MAJOR.MINOR.PATCH. The basic intent is that MAJOR versions are
>incompatible, large-scale upgrades of the API. MINOR versions
>retain usage and extension compatibility with older minor versions,
>and changes in the PATCH level are perfectly compatible, forwards
>and backwards.
>
>Following the main design principle of Cocoon, the pyramid
>of contracts, we distinguish between users and developers
>of Cocoon.
>A very rough distinction between them is that a user writes
>the application using Cocoon without coding Java. There is
>one exception to this rule: flow script - the java script
>is also written by the user.
>A developer codes Java and provides additional functionality
>for the user, so a developer extends Cocoon.
>
>Therefore we distinguish between usage compatibility and
>extension compatibility. Both compatibility levels cover
>some kind of "source" compatibility. Cocoon does not provide
>binary compatibility. But as Cocoon is distributed as a
>source release that you have to compile anyway, it's
>saver to compile your own application code (if any) using
>the version of Cocoon that your application runs on.
>
>Usage Compatibility
>-------------------
>
>'Usage' compatibility guarantees that an application written
>by a Cocoon user is compatible. All files developed by a typical
>Cocoon user like xml files, sitemaps, stylesheets (elements and
>namespace declarations) keep on being picked up by the machinery
>and are dealt with correctly (sitemap semantics,
>generator/transformer-picked up elements, config file entries...).
>In fact this should cover everything (including flow script) but
>except own Java code.
>
>Applications that write against a particular version will remain
>usage compatible against later versions, until the major number
>changes.
>Writing an application against a version means that this application
>does not use any deprecated API of that version. Therefore minor
>version changes are only usage compatible from one minor version to
>the direct following one. For example 2.2 is usage compatible to 2.1.
>But 2.3 is not necessary usage compatible to 2.1, although it is
>usage compatible to 2.2. As long as you don't use deprecated API,
>your application is usage compatible across all minor versions.
>
>Example:
>- a feature is introduced in 2.0 and used by the application.
>- it is deprecated in 2.2.
>- it will be removed in 2.3.
>
>However, if an application uses an API which has become available
>in a particular minor version, it (obviously) will no longer
>operate against previous minor versions.
>
>Extension Compatibility
>-----------------------
>
>'extension' compatibility guarantees that own extensions to what
>Cocoon provides (own Java classes that interface directly with
>API in the Cocoon distribution) compile and operate.
>
>Applications that write against a particular version will remain
>extension compatible against later versions until the major or
>the minor number changes (Please note the difference to the usage
>compatibility). However, the Cocoon developers take care that even
>if the minor number changes, most of the own code still works and
>operates properly. Incompatible changes between minor versions are
>kept to a minimum. Frequent new releases of Cocoon ensure that
>developers have a smooth transition path.
>
>If an interface/class changes in an incompatible way between minor
>version changes, the Cocoon documentation will contain a detailed
>update guide that contains a solution for every incompatible change.
>So following this guide should make the update smoothly.
>
>
>Deprecation and Exceptions
>--------------------------
>
>To continue the Cocoon development and to keep up with the innovations,
>parts of Cocoon might get deprecated; this includes parts of the user
>API and also parts of the developer API.
>
>If a part of the user API is deprecated, this will be flagged through
>run-time warnings that appear in the logs but remain supported. This
>indicates that an upcoming minor (or major) release will no longer
>support this.
>
>If a part of the developer API is deprecated it will be removed with
>the next major, minor or patch release.
>
>For developers there is one exception to this rule: private API. Cocoon
>has some internal classes and interfaces that are not meant to be used
>by a Cocoon developer (someone extending Cocoon). These pieces of
>Java code are clearly marked in the Javadocs and should not be used.
>They might change even between a patch version change in an incompatible
>way without providing a workaround!
>
>External Libraries
>------------------
>Cocoon uses a set of external libraries (like for example Avalon,
>Xalan or Xerces). Inbetween any release, even patch releases,
>the versions of the external libraries might be updated to any version.
>
>Therefore if your application is written against a special API of an
>external library it might be that this API of the external library
>changes inbetween two Cocoon versions and therefore your application
>does not work properly anymore (or even does not compile anymore).
>Unfortunately, this issue is out of the scope of Cocoon.
>
>
>Examples
>--------
>Here are some examples to demonstrate the compatibility:
>
>Original Version New Version Usage Compatible Extension Compatible
>2.2.3 2.2.4 Yes Yes
>2.2.3 2.3.1 Yes No
>2.2.3 3.0.0 No No
>
>Note: while some of the cells say "no", it is possible that the versions
>may be compatible, depending very precisely upon the particular APIs
>used by the application.
>
>
>Versioning and Repositories
>---------------------------
>Cocoon is very innovative and new features are added very frequently. This
>results in new functionality that justify a minor version change on its own.
>In addition, to move Cocoon forward technologically, some features have
>to be removed/deprecated which of course results in a minor version change
>as well.
>
>So, Cocoon will potentially release several versions with minor version
>changes in a year and only a few patch releases. Of course, if a patch
>release is required, for example if a potential security issue is found
>etc., the patch release will be made available asap.
>
>This high innovention has - at least in theory - the price of maintaining
>several branches at once in order to be able to apply patches to older
>versions.
>To reduce this to a minimum, Cocoon uses one repository per major version.
>The HEAD of the repository always contains the latest version. If a new
>minor release is required, the version in the CVS is directly changed
>to this minor version (e.g. from 2.1.4 to 2.2). The current state is
>tagged and if the need for a 2.1.5 release arises, a branch is created.
>
>However, due to the compatibility rules explained above, there should only
>be a reason for a patch version if major problems occur (like security
>issues). In any other case, the Cocoon community expects the applications
>to migrate to the new minor version release which should be fairly simple.
>
>In addition Cocoon has a sandbox repository to test new features.
>
>
>Carsten
>
>Carsten Ziegeler
>Open Source Group, S&N AG
>http://www.osoco.net/weblogs/rael/
>
>
>
>
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Bruno Dumon <br...@outerthought.org>.
On Wed, 2004-04-21 at 09:33, Carsten Ziegeler wrote:
<snip/>
> External Libraries
> ------------------
> Cocoon uses a set of external libraries (like for example Avalon,
> Xalan or Xerces). Inbetween any release, even patch releases,
> the versions of the external libraries might be updated to any version.
If between patch releases we want to keep everything as compatible as
possible, then we shouldn't upgrade external libraries either (except if
known to be completely safe). Otherwise, all the effort we put into
keeping our own stuff compatible could be lost because of that.
>
> Therefore if your application is written against a special API of an
> external library it might be that this API of the external library
> changes inbetween two Cocoon versions and therefore your application
> does not work properly anymore (or even does not compile anymore).
> Unfortunately, this issue is out of the scope of Cocoon.
--
Bruno Dumon http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
bruno@outerthought.org bruno@apache.org
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Marc Portier wrote:
>
> I currently don't think we have a release scheme that
> supports one or the other: i.e. reality seems pretty much
> like what Carsten is saying:
> we just have 1,2,... 1004 (based on some gut feeling we seem
> to be distributing those numbers over tripplets)
>
I thought about this topic a little bit and I can only say we
never followed any compatibility rules! For example we changed
the sitemap in an incompatible way not only from 2.0.x to 2.1,
but also from 2.1.3 to 2.1.4.
And we removed deprecated API/changed interfaces/methods
between patch releases.
So following this "logic", we could just follow this road and do
whatever we think is appropriate or establish some general and
simple rules to make the live of the users easier.
I have written a first draft of such a guide (well I copied some
things from here and there and added my two cents). So, let's
start another flame war and tell me what you think about this.
(PS: I think this one of the longest mails I ever wrote)
The Cocoon Versioning Manifest (CVM)
------------------------------------
This document covers how the Cocoon project is versioned.
Since Cocoon is a framework, it is very important to define
a stable API for users and developers. However, we also need
to move Cocoon forward, technologically. To balance these two
needs, a strict policy of versioning is required, which users
can rely upon to understand the limitations, restrictions,
and the changes that can occur from one release to the next.
Basics
------
Versions are denoted using a standard triplet of integers:
MAJOR.MINOR.PATCH. The basic intent is that MAJOR versions are
incompatible, large-scale upgrades of the API. MINOR versions
retain usage and extension compatibility with older minor versions,
and changes in the PATCH level are perfectly compatible, forwards
and backwards.
Following the main design principle of Cocoon, the pyramid
of contracts, we distinguish between users and developers
of Cocoon.
A very rough distinction between them is that a user writes
the application using Cocoon without coding Java. There is
one exception to this rule: flow script - the java script
is also written by the user.
A developer codes Java and provides additional functionality
for the user, so a developer extends Cocoon.
Therefore we distinguish between usage compatibility and
extension compatibility. Both compatibility levels cover
some kind of "source" compatibility. Cocoon does not provide
binary compatibility. But as Cocoon is distributed as a
source release that you have to compile anyway, it's
saver to compile your own application code (if any) using
the version of Cocoon that your application runs on.
Usage Compatibility
-------------------
'Usage' compatibility guarantees that an application written
by a Cocoon user is compatible. All files developed by a typical
Cocoon user like xml files, sitemaps, stylesheets (elements and
namespace declarations) keep on being picked up by the machinery
and are dealt with correctly (sitemap semantics,
generator/transformer-picked up elements, config file entries...).
In fact this should cover everything (including flow script) but
except own Java code.
Applications that write against a particular version will remain
usage compatible against later versions, until the major number
changes.
Writing an application against a version means that this application
does not use any deprecated API of that version. Therefore minor
version changes are only usage compatible from one minor version to
the direct following one. For example 2.2 is usage compatible to 2.1.
But 2.3 is not necessary usage compatible to 2.1, although it is
usage compatible to 2.2. As long as you don't use deprecated API,
your application is usage compatible across all minor versions.
Example:
- a feature is introduced in 2.0 and used by the application.
- it is deprecated in 2.2.
- it will be removed in 2.3.
However, if an application uses an API which has become available
in a particular minor version, it (obviously) will no longer
operate against previous minor versions.
Extension Compatibility
-----------------------
'extension' compatibility guarantees that own extensions to what
Cocoon provides (own Java classes that interface directly with
API in the Cocoon distribution) compile and operate.
Applications that write against a particular version will remain
extension compatible against later versions until the major or
the minor number changes (Please note the difference to the usage
compatibility). However, the Cocoon developers take care that even
if the minor number changes, most of the own code still works and
operates properly. Incompatible changes between minor versions are
kept to a minimum. Frequent new releases of Cocoon ensure that
developers have a smooth transition path.
If an interface/class changes in an incompatible way between minor
version changes, the Cocoon documentation will contain a detailed
update guide that contains a solution for every incompatible change.
So following this guide should make the update smoothly.
Deprecation and Exceptions
--------------------------
To continue the Cocoon development and to keep up with the innovations,
parts of Cocoon might get deprecated; this includes parts of the user
API and also parts of the developer API.
If a part of the user API is deprecated, this will be flagged through
run-time warnings that appear in the logs but remain supported. This
indicates that an upcoming minor (or major) release will no longer
support this.
If a part of the developer API is deprecated it will be removed with
the next major, minor or patch release.
For developers there is one exception to this rule: private API. Cocoon
has some internal classes and interfaces that are not meant to be used
by a Cocoon developer (someone extending Cocoon). These pieces of
Java code are clearly marked in the Javadocs and should not be used.
They might change even between a patch version change in an incompatible
way without providing a workaround!
External Libraries
------------------
Cocoon uses a set of external libraries (like for example Avalon,
Xalan or Xerces). Inbetween any release, even patch releases,
the versions of the external libraries might be updated to any version.
Therefore if your application is written against a special API of an
external library it might be that this API of the external library
changes inbetween two Cocoon versions and therefore your application
does not work properly anymore (or even does not compile anymore).
Unfortunately, this issue is out of the scope of Cocoon.
Examples
--------
Here are some examples to demonstrate the compatibility:
Original Version New Version Usage Compatible Extension Compatible
2.2.3 2.2.4 Yes Yes
2.2.3 2.3.1 Yes No
2.2.3 3.0.0 No No
Note: while some of the cells say "no", it is possible that the versions
may be compatible, depending very precisely upon the particular APIs
used by the application.
Versioning and Repositories
---------------------------
Cocoon is very innovative and new features are added very frequently. This
results in new functionality that justify a minor version change on its own.
In addition, to move Cocoon forward technologically, some features have
to be removed/deprecated which of course results in a minor version change
as well.
So, Cocoon will potentially release several versions with minor version
changes in a year and only a few patch releases. Of course, if a patch
release is required, for example if a potential security issue is found
etc., the patch release will be made available asap.
This high innovention has - at least in theory - the price of maintaining
several branches at once in order to be able to apply patches to older
versions.
To reduce this to a minimum, Cocoon uses one repository per major version.
The HEAD of the repository always contains the latest version. If a new
minor release is required, the version in the CVS is directly changed
to this minor version (e.g. from 2.1.4 to 2.2). The current state is
tagged and if the need for a 2.1.5 release arises, a branch is created.
However, due to the compatibility rules explained above, there should only
be a reason for a patch version if major problems occur (like security
issues). In any other case, the Cocoon community expects the applications
to migrate to the new minor version release which should be fairly simple.
In addition Cocoon has a sandbox repository to test new features.
Carsten
Carsten Ziegeler
Open Source Group, S&N AG
http://www.osoco.net/weblogs/rael/
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Marc Portier wrote:
>
> +1
>
> - one extra: when the external lib mismatch can occur, then
> we should provide that knowledge in the update docs and add a warning
Yepp.
>
> -marc= (being glad he posted that arp link :-))
Yes, and I also stole text from your mail :)
>
> <snip
> what="the historical long mail from one eloquent Sir C. Ziegler" />
ROTFL
Carsten
Re: Proposal: release management guide (was Re: [RT] Versions)
Posted by Marc Portier <mp...@outerthought.org>.
+1
- one extra: when the external lib mismatch can occur, then we should
provide that knowledge in the update docs and add a warning
-marc= (being glad he posted that arp link :-))
<snip
what="the historical long mail from one eloquent Sir C. Ziegler" />
--
Marc Portier http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
Read my weblog at http://blogs.cocoondev.org/mpo/
mpo@outerthought.org mpo@apache.org
RE: Proposal: release management guide (was Re: [RT] Versions)
Posted by Carsten Ziegeler <cz...@s-und-n.de>.
Marc Portier wrote:
> here goes:
> I see some difference between what I would call 'extension'
> vs. 'usage'
> compatibility
>
> <snip explains/>
I like the distinction and I totally agree with the vision of
'usage'. In my RT about versions I was only refering to
the 'extension' are when it comes to incompatible changes.
>
> I currently don't think we have a release scheme that
> supports one or the other: i.e. reality seems pretty much
> like what Carsten is saying:
> we just have 1,2,... 1004 (based on some gut feeling we seem
> to be distributing those numbers over tripplets)
>
Yes, I think this is really the truth. And in reality things
get very complicated as compatibility for Cocoon as a whole
applies not only to what we develop but to most other projects
we incorporate.
We are - for example - very eager in updating to new releases
of external libraries. So it happens, that we change from
Xalan 2.4 to Xalan 2.6 between 2.1.3 and 2.1.4. (just a fictional
sample). It's a major change for Xalan, so it's not required
to be compatible. But if Cocoon or a users application relies
on some things that have changed between 2.4 and 2.6, then
Cocoon isn't compatible between 2.1.3 and 2.1.4 for them. And
of course, they blame Cocoon (which isn't soo wrong).
And in fact, these updates to the external libraries cause a
lot of problems to users when they try to update from one
version to another of Cocoon.
So, what do we do with all these infos?
Carsten