You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Ryan Bloom <rb...@ntrnet.net> on 2000/03/20 17:05:40 UTC
A couple of important features for 2.0
I have had a couple of discussions with people recently, and I would like
to add a couple of items to the todo list for 2.0. I think if we can get
this stuff into the release ASAP, we can still stick to our current
release schedule.
1) IO Layering. The enabling logic has been in for a few months, we need
to turn it on. There are multiple groups of people who need this. I
would like to turn this one this week for the next alpha (which I would
like to release Friday, more on this in the next note).
2) Backwards compatability. It would be VERY nice if we could figure out
some way to make modules backwards compatable within major releases. I
don't mind breaking compatability between 2.0 and 3.0, but between 2.0 and
2.1, modules shouldn't need to be re-compiled. :-) I have no good ideas
on how to do this, but I will be working on it.
3) Inter-module communication. Basically, what EAPI does for 1.3. I
need to look into what this means in much more detail. I'll post more
when I understand the request. (I'm not sure the IOL doesn't handle this,
but I'll look into it and post more ASAP).
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by rb...@apache.org.
I've been looking at this. As far as testing the capability of make to
find the files correctly, that is relatively easy using AP_TRY_RUN. I
have been trying to figure out the build scripts, but I haven't really had
the time to devote to them recently. I personally think the Apache
configuration for 2.0 is way too complicated. I can't figure out why we
are using APR if Apache is going to check everything too. I would like to
see us using absolute paths in the makefiles, but I haven't looked into
how to do it yet.
Ryan
On Mon, 20 Mar 2000, Ben Laurie wrote:
> Ryan Bloom wrote:
> >
> > I have had a couple of discussions with people recently, and I would like
> > to add a couple of items to the todo list for 2.0. I think if we can get
> > this stuff into the release ASAP, we can still stick to our current
> > release schedule.
>
> Oh, and I'd really like an answer to my query about make and paths on
> Tru64. Or I'm gonna start hacking randomly :-)
>
> Cheers,
>
> Ben.
>
> --
> SECURE HOSTING AT THE BUNKER: http://www.thebunker.net/hosting.htm
>
> http://www.apache-ssl.org/ben.html
>
> Coming to ApacheCon Europe? http://ApacheCon.Com/
>
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by Ben Laurie <be...@algroup.co.uk>.
Ryan Bloom wrote:
>
> I have had a couple of discussions with people recently, and I would like
> to add a couple of items to the todo list for 2.0. I think if we can get
> this stuff into the release ASAP, we can still stick to our current
> release schedule.
Oh, and I'd really like an answer to my query about make and paths on
Tru64. Or I'm gonna start hacking randomly :-)
Cheers,
Ben.
--
SECURE HOSTING AT THE BUNKER: http://www.thebunker.net/hosting.htm
http://www.apache-ssl.org/ben.html
Coming to ApacheCon Europe? http://ApacheCon.Com/
Re: A couple of important features for 2.0
Posted by rb...@apache.org.
On Mon, 20 Mar 2000, Ben Laurie wrote:
> Ryan Bloom wrote:
> > 3) Inter-module communication. Basically, what EAPI does for 1.3. I
> > need to look into what this means in much more detail. I'll post more
> > when I understand the request. (I'm not sure the IOL doesn't handle this,
> > but I'll look into it and post more ASAP).
>
> I'd also like to hear the background to the request. One concern I have
> with this is that full generality will lead to an undisciplined,
> unmaintainable mess. It seems far more sensible to me to figure out what
> modules want to use this for and add more hooks to the core to enable
> them to do it in a disciplined and module-independent way.
That's why I'm doing more research before commiting to this at all. When
I figure out why this is required and if their is another way to
accomplish it.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by Ben Laurie <be...@algroup.co.uk>.
Ryan Bloom wrote:
> 3) Inter-module communication. Basically, what EAPI does for 1.3. I
> need to look into what this means in much more detail. I'll post more
> when I understand the request. (I'm not sure the IOL doesn't handle this,
> but I'll look into it and post more ASAP).
I'd also like to hear the background to the request. One concern I have
with this is that full generality will lead to an undisciplined,
unmaintainable mess. It seems far more sensible to me to figure out what
modules want to use this for and add more hooks to the core to enable
them to do it in a disciplined and module-independent way.
BTW +1 on IO Layering.
Cheers,
Ben.
--
SECURE HOSTING AT THE BUNKER: http://www.thebunker.net/hosting.htm
http://www.apache-ssl.org/ben.html
Coming to ApacheCon Europe? http://ApacheCon.Com/
Re: A couple of important features for 2.0
Posted by Dean Gaudet <dg...@arctic.org>.
On Mon, 20 Mar 2000, Ryan Bloom wrote:
> 2) Backwards compatability. It would be VERY nice if we could figure out
> some way to make modules backwards compatable within major releases. I
> don't mind breaking compatability between 2.0 and 3.0, but between 2.0 and
> 2.1, modules shouldn't need to be re-compiled. :-) I have no good ideas
> on how to do this, but I will be working on it.
please don't try this. please please not. you can stifle performance
improvements if you lock down the API at the binary level.
Dean
Re: A couple of important features for 2.0
Posted by rb...@apache.org.
On Fri, 24 Mar 2000, Manoj Kasichainula wrote:
> On Fri, Mar 24, 2000 at 01:30:21AM +0000, Tony Finch wrote:
> > Maybe I'm being a dolt but I don't understand what you mean by a
> > "fundamental change"
>
> Sometimes, a change is so thorough that you can't really provide
> backwards compatibility, because the old API was just not designed to
> support such a radically different infrastructure.
That's true, but why is that change going into a point release?
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by Manoj Kasichainula <ma...@io.com>.
On Fri, Mar 24, 2000 at 01:30:21AM +0000, Tony Finch wrote:
> Maybe I'm being a dolt but I don't understand what you mean by a
> "fundamental change"
Sometimes, a change is so thorough that you can't really provide
backwards compatibility, because the old API was just not designed to
support such a radically different infrastructure.
Re: A couple of important features for 2.0
Posted by Tony Finch <do...@dotat.at>.
Dean Gaudet <dg...@arctic.org> wrote:
>On Tue, 21 Mar 2000, Tony Finch wrote:
>
>> This doesn't solve the problem of changing data structures; I think
>> that will have to be dealt with by adding accessor functions to the
>> API and making third party modules use them.
>
>also doesn't solve fundamental changes... picking one out of the 1.3
>CHANGES file at random:
>
> *) API: A new field in the request_rec, r->mtime, has been added to
> avoid gratuitous parsing of date strings.
Maybe I'm being a dolt but I don't understand what you mean by a
"fundamental change" and in particular I don't see what the problem is
with this example change.
Tony.
--
f.a.n.finch fanf@demon.net dot@dotat.at
268 the unmistakable splatter of rotten mellons
Re: A couple of important features for 2.0
Posted by Dean Gaudet <dg...@arctic.org>.
On Tue, 21 Mar 2000, Tony Finch wrote:
> This doesn't solve the problem of changing data structures; I think
> that will have to be dealt with by adding accessor functions to the
> API and making third party modules use them.
also doesn't solve fundamental changes... picking one out of the 1.3
CHANGES file at random:
*) API: A new field in the request_rec, r->mtime, has been added to
avoid gratuitous parsing of date strings. It is intended to hold
the last-modified date of the resource (if applicable). An
update_mtime() routine has also been added to advance it if
appropriate. [Roy Fielding, Ken Coar]
Dean
Re: A couple of important features for 2.0
Posted by Tony Finch <do...@dotat.at>.
"Bill Stoddard" <st...@raleigh.ibm.com> wrote:
>>
>> 2) Freeze the current API. When a new feature is required, add a new
>> function instead of modifying the old one.
>
>This with 1) helps even more. Bloats the code, stifles creativity and just
>doesn't fit in with open source development philosophy. If an API is just
>not right, it should be changed. Maintaining deprecated APIs is common
>practice in commercial software development but there is a price to pay in
>complexity. And complexity is an open source project killer, IMHO; it raises
>the project entry barrier and bugs increase non-linearly with complexity.
These are the issues I was trying to address with the API jump table
but I really don't like that idea for a number of reasons (mostly
variations on the theme of run-time complexity).
I came up with an alternative this evening which avoids the
performance problems of using a jump table by using pre-processor
magic instead. To #include the Apache API a module says something like
#define AP_API_VERSION 20000320
#include "ap_api.h"
and then proceeds to program to the API without worrying about
changing versions -- source compatibility and binary compatibility.
The core is programmed in a similar way, except it always uses the
most recent API version, e.g. all core code says
#include "ap_api_latest.h"
#include "ap_api.h"
before doing anything else. However, when an API change is made, the
old version of the function that has been changed is retained
(probably as a wrapper calling the new version). Any old versions are
distinguished from the new one by mangling their names with different
date strings, e.g.
ap_foo_19991207
ap_foo_20000319
ap_foo
and the latest version of the function has an unmangled name.
All unmangled names are really mangled names in disguise: the
"ap_api.h" header expands them to whatever is appropriate, according
to the value of AP_API_VERSION. "ap_api.h" knows when each API
function was changed, and when #included does
#define ap_foo ap_foo_XXXXXXXX
for the largest XXXXXX not greater than AP_API_VERSION (so if I say
#define AP_API_VERSION 20000202
#include "ap_api.h"
then when I use ap_foo I'll actually be using ap_foo_19991207). A
little tool will be required to maintain "ap_api_latest.h" and
"ap_api.h", using a source file with content similar to what is in
"ap_mmn.h" right now (i.e. I don't expect the maintenance burden to be
a great deal worse).
This doesn't solve the problem of changing data structures; I think
that will have to be dealt with by adding accessor functions to the
API and making third party modules use them.
Tony.
--
f.a.n.finch fanf@demon.net dot@dotat.at
369 the barnacle-encrusted bilge pump on your sunken dreams
Re: A couple of important features for 2.0
Posted by Bill Stoddard <st...@raleigh.ibm.com>.
>
> I disagree. This doesn't stifle creativity at all, and it doesn't
> increase complexity that much. If a function was useful in one version,
> there was a reason. If it is later found to not be useful enough, it
> should be changed, but the question is how to change it. You could add
> code to the current function (increasing complexity for that function), or
> you could add a new function which uses the old function. The second
> option maintains backwards compatability and modularizes the code.
> Explain how this increases complexity.
Ever heard of the old saying about why God was able to create the world in 7
days? He didn't have an installed base to reckon with.
My experience with maintaining seamless backward compatability with an
Apache style 'API' is that it is hard and it will be error prone. You have
to do it pretty much the way you've outlined; add padding to the end of
structures, deprecate existing APIs (but leave them in, even if they don't
quite function the same because of architecture changes) in favor of newer
APIs, etc. And some of the errors will be subtle as hell; a module that
works with v1 works fine for three weeks on v1.1 then fails unexpectedly
with a core dump. I've spent weeks debugging customer problems like this. If
we are going to advertise that we have binary backward compatability, I want
to make damn sure it is true. I am wary of any solution that attempts to
make the Apache API 'hardened'. It was not designed for that.
Bill
Re: A couple of important features for 2.0
Posted by Dean Gaudet <dg...@arctic.org>.
On Mon, 20 Mar 2000 rbb@apache.org wrote:
> I disagree. This doesn't stifle creativity at all, and it doesn't
> increase complexity that much.
you should track linux device driver API changes over the history of
linux, it's a really similar problem. each time the API was changed (at
the source level even) it was for a desirable feature -- such as cleaner
SMP support, better performance, or improved features at a higher level
(such as improvements in the generic networking layer causing changes to
the underlying drivers).
the problem is a well studied one, and you'd have to do a lot more than
just hand waving to convince me that you've got something new and
wonderful to handle it.
if you do have something new and wonderful, let's start doing it that way
in 2.1 and not delay 2.0 waiting for a demonstration of your new technique
:)
Dean
Re: A couple of important features for 2.0
Posted by rb...@apache.org.
On Mon, 20 Mar 2000, Bill Stoddard wrote:
>
> > Well, there are a couple of things we can do to try to keep modules
> > backwards compatable.
> >
> > 1) Add some void space at the end of internal structures to allow them to
> > grow without breaking binary compatability.
> This would help but is not in itself a solution.
No, it isn't a solution, but it does solve a problem that we have. For
example, you added a field to the request_rec earlier today, if we had
space at the end of the structure, we wouldn't have had to bump the MMN.
> > 2) Freeze the current API. When a new feature is required, add a new
> > function instead of modifying the old one.
> This with 1) helps even more. Bloats the code, stifles creativity and just
> doesn't fit in with open source development philosophy. If an API is just
> not right, it should be changed. Maintaining deprecated APIs is common
> practice in commercial software development but there is a price to pay in
> complexity. And complexity is an open source project killer, IMHO; it raises
> the project entry barrier and bugs increase non-linearly with complexity.
I disagree. This doesn't stifle creativity at all, and it doesn't
increase complexity that much. If a function was useful in one version,
there was a reason. If it is later found to not be useful enough, it
should be changed, but the question is how to change it. You could add
code to the current function (increasing complexity for that function), or
you could add a new function which uses the old function. The second
option maintains backwards compatability and modularizes the code.
Explain how this increases complexity.
As far as bloating the code goes, I disagree with this as well. The
function was required at one time. This tells me that there are modules
using the original function. Adding features to that function that
require the prototype to change means bloating the modules that are
currently using the function.
Ryan
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by Bill Stoddard <st...@raleigh.ibm.com>.
> Well, there are a couple of things we can do to try to keep modules
> backwards compatable.
>
> 1) Add some void space at the end of internal structures to allow them to
> grow without breaking binary compatability.
This would help but is not in itself a solution.
>
> 2) Freeze the current API. When a new feature is required, add a new
> function instead of modifying the old one.
This with 1) helps even more. Bloats the code, stifles creativity and just
doesn't fit in with open source development philosophy. If an API is just
not right, it should be changed. Maintaining deprecated APIs is common
practice in commercial software development but there is a price to pay in
complexity. And complexity is an open source project killer, IMHO; it raises
the project entry barrier and bugs increase non-linearly with complexity.
>
> 3) Try to make the API dynamic. (I'm not sure if this is even feasible).
> Basically, the module registers the level of the API that it expects, and
> we try to accomodate this.
Hummm... Looking forward to your thoughts...
>
> I'm still looking into this issue.
Why not create a stable API designed for those who cannot or do not wish to
recompile modules (ISAPI, NSAPI, et. al)? It will not be as functional, but
maybe we could capture 90% of the functionality with stable function
wrappers. And of course this API would NOT allow frobbing internal
structures directly. I think Rasmus was, at one time, interested in this.
All of these options are less than satisfying. Bleh.
Bill
Re: A couple of important features for 2.0
Posted by rb...@apache.org.
Well, there are a couple of things we can do to try to keep modules
backwards compatable.
1) Add some void space at the end of internal structures to allow them to
grow without breaking binary compatability.
2) Freeze the current API. When a new feature is required, add a new
function instead of modifying the old one.
3) Try to make the API dynamic. (I'm not sure if this is even feasible).
Basically, the module registers the level of the API that it expects, and
we try to accomodate this.
I'm still looking into this issue.
Ryan
On Mon, 20 Mar 2000, Tony Finch wrote:
> Ryan Bloom <rb...@ntrnet.net> wrote:
> >
> >2) Backwards compatability. It would be VERY nice if we could figure out
> >some way to make modules backwards compatable within major releases.
>
> The only way I can think of doing this reliably (other than by
> freezing the API) is for modules to make all API calls via a vector of
> function pointers which is initialised when the module starts and
> tells the core what version of the API it is written to.
>
> Tony.
> --
> f.a.n.finch fanf@demon.net dot@dotat.at
> 320 the dimpled rind of remorselessness
>
_______________________________________________________________________________
Ryan Bloom rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
Re: A couple of important features for 2.0
Posted by Tony Finch <do...@dotat.at>.
Ryan Bloom <rb...@ntrnet.net> wrote:
>
>2) Backwards compatability. It would be VERY nice if we could figure out
>some way to make modules backwards compatable within major releases.
The only way I can think of doing this reliably (other than by
freezing the API) is for modules to make all API calls via a vector of
function pointers which is initialised when the module starts and
tells the core what version of the API it is written to.
Tony.
--
f.a.n.finch fanf@demon.net dot@dotat.at
320 the dimpled rind of remorselessness