You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Max Bowsher <ma...@ukf.net> on 2005/04/05 16:05:28 UTC
Very emphatic -1 for r13369 (svn:keywords canonicalization)
r13369 canonicalizes svn:keywords values.
The canonicalization algorithm is flawed, and silently deletes anything that
it does not understand.
This means that when we add additional keywords in the future, old clients
will be unable to edit a svn:keywords property containing a new value
without inadvertently removing all the keywords which the old client version
does not support.
Also, we have an issue regarding custom keywords. I'm not sure if this has
been considered yet, but if it is to extend the svn:keywords syntax in any
way, it is vital that we do not release a version with a destructive
canonicalization routine.
On this basis, I am about to revert the revision in question, and call a
vote in the 1.2.x STATUS file for reverting it there too.
Once we have cleared up these issues, we can reintroduce a properly designed
canonicalization system in 1.3.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Madan U Sreenivasan <ma...@collab.net>.
On Tue, 2005-04-05 at 21:35, Max Bowsher wrote:
> r13369 canonicalizes svn:keywords values.
>
> The canonicalization algorithm is flawed, and silently deletes anything that
> it does not understand.
>
> This means that when we add additional keywords in the future, old clients
> will be unable to edit a svn:keywords property containing a new value
> without inadvertently removing all the keywords which the old client version
> does not support.
>
I thought this would be a good fallout of the canonicalization ( removal
of junk input ). However, I agree with your point.
[...]
> Once we have cleared up these issues, we can reintroduce a properly designed
> canonicalization system in 1.3.
>
I have sent a patch ( the whole patch ) with modified test cases. Pl.
let me know if this satisfies the above point. Thanks for the review.
>
> Max.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by John Peacock <jp...@rowman.com>.
Max Bowsher wrote:
> The canonicalization algorithm is flawed, and silently deletes anything
> that it does not understand.
Very good catch! It's a pity you didn't notice it before 1.2 branched, because
it is actually relatively easy to fix: canonicalize the keywords that are
already defined and just pass through as is anything that isn't recognized.
It's a little harder to ensure that only a single alias of a given keyword is
present, but not too hard.
Since you pulled r13369, you also need to pull r1203 in the book repository,
which was the documentation for the new behavior...
John
--
John Peacock
Director of Information Research and Technology
Rowman & Littlefield Publishing Group
4720 Boston Way
Lanham, MD 20706
301-459-3366 x.5010
fax 301-429-5747
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by kf...@collab.net.
"Max Bowsher" <ma...@ukf.net> writes:
> r13369 canonicalizes svn:keywords values.
>
> The canonicalization algorithm is flawed, and silently deletes
> anything that it does not understand.
>
> This means that when we add additional keywords in the future, old
> clients will be unable to edit a svn:keywords property containing a
> new value without inadvertently removing all the keywords which the
> old client version does not support.
>
> Also, we have an issue regarding custom keywords. I'm not sure if this
> has been considered yet, but if it is to extend the svn:keywords
> syntax in any way, it is vital that we do not release a version with a
> destructive canonicalization routine.
>
> On this basis, I am about to revert the revision in question, and call
> a vote in the 1.2.x STATUS file for reverting it there too.
+1 on your -1. Glad you caught this (and embarrassed for not thinking
of it myself!).
> Once we have cleared up these issues, we can reintroduce a properly
> designed canonicalization system in 1.3.
Agreed.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Max Bowsher <ma...@ukf.net>.
Greg Hudson wrote:
> On Wed, 2005-04-06 at 15:07, kfogel@collab.net wrote:
>> Just FYI, I've never felt that way, since it's as easy to unrevert+fix
>> the reversion that someone else did. It's all under version control,
>> therefore it's all easy :-).
>
> It does tend to make it harder to get good svn blame information on a
> file if it's been through a lot of reversions and unreversions. They
> can make you have to blame a lot of historical revisions of the file to
> find out who actually wrote a line of code.
>
> We shouldn't be so attached to blame as to avoid all code
> reorganizations and reversions, but we shouldn't assign a zero cost to
> such things either.
Happily, in this case, the change was almost exclusively the addition, not
changing, of lines, so there is no muddying of blame info.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Greg Hudson <gh...@MIT.EDU>.
On Wed, 2005-04-06 at 15:07, kfogel@collab.net wrote:
> Just FYI, I've never felt that way, since it's as easy to unrevert+fix
> the reversion that someone else did. It's all under version control,
> therefore it's all easy :-).
It does tend to make it harder to get good svn blame information on a
file if it's been through a lot of reversions and unreversions. They
can make you have to blame a lot of historical revisions of the file to
find out who actually wrote a line of code.
We shouldn't be so attached to blame as to avoid all code
reorganizations and reversions, but we shouldn't assign a zero cost to
such things either.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Max Bowsher <ma...@ukf.net>.
Greg Stein wrote:
> On Wed, Apr 06, 2005 at 02:07:43PM -0500, kfogel@collab.net wrote:
>> Julian Foad <ju...@btopenworld.com> writes:
>>> Max Bowsher wrote:
>>>> On this basis, I am about to revert the revision in question, and
>>>> call a vote in the 1.2.x STATUS file for reverting it there too.
>>>
>>> By the way, it was me that committed that patch on behalf of another
>>> contributor, and I would have been happy to revert it myself if you
>>> had just pointed out this problem and suggested reverting it. It is
>>> usually considered rude to revert someone else's work without first
>>> giving them the chance to fix it. Don't worry about it this time,
>>> though - I'm cool :-)
>>
>> Just FYI, I've never felt that way, since it's as easy to unrevert+fix
>> the reversion that someone else did. It's all under version control,
>> therefore it's all easy :-).
>>
>> I think Max just wanted to act quickly due to 1.2 scheduling.
>
> I concur with Julian. Reverting somebody else's work is (IMO) bad form.
Certainly it can be, but any bad form comes from the attititude expressed
along with the reversion, not the reversion itself.
> Recognize that a very likely outcome would be to fix the change rather
> than to back it out. By stepping in and reverting the change you are
> essentially saying, "I'm vetoing this and there is nothing you can do
> about it. There is no other resolution we can come to. Don't even
> bother to ask." It completely shuts down any discussion about the
> change.
No, no it doesn't. It simply says "not having this commit is better than
having it, so let's make an incremental improvement by removing it.". I
explained what the problems were with the commit: One unequivocal real bug,
and one point where the design has been incompletely thought through.
> A veto is supposed to mean, "I disagree with this change. Let's talk
> about it." Stuff is not allowed to ship with vetoed code in it, sure,
> but there should be an assumption of *discussion* rather than
> unilateral action.
Exactly. This wasn't just an "I don't like this", it was a "look, there is a
real demonstrable and quite unpleasant bug". Thus, I took steps to get the
bug removed, so it wouldn't hinder the progression of 1.2, and suggested we
proceed with discussion to sort out how to do this properly. Granted, I
assumed that the fixed code would be placed in 1.3, not 1.2, but had no evil
motivation - I was just taking note that it was a new feature, and 1.2 is
branched, and we do not usually backport features.
> I'm not sure how to stress this any more. Let the original committer
> revert the change, or fix it. A vetoer should never do that.
If there is a clear problem, not merely a difference of opinions, I really
don't see why the above matters at all. The importance is all in the
justification of the action, not who takes it, for me.
> If there is a time pressure due to a release, then the *RM* can
> certainly revert the thing in the release branch (or possibly on trunk
> if that's where the release is coming from; tho I'd recommend a
> branch, a reversion, and then release). The RM is free to take this
> kind of action. Developers should not; this isn't about "easy" or
> "hard" but about social dynamics and the expectation of discussion.
I basically said "This has problems {explanation} .... Let's discuss!"
I certainly do expect discussion!
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by kf...@collab.net.
Greg Stein <gs...@lyra.org> writes:
> I concur with Julian. Reverting somebody else's work is (IMO) bad form.
>
> Recognize that a very likely outcome would be to fix the change rather
> than to back it out. By stepping in and reverting the change you are
> essentially saying, "I'm vetoing this and there is nothing you can do
> about it. There is no other resolution we can come to. Don't even
> bother to ask." It completely shuts down any discussion about the
> change.
I don't think that message is really inherent in a reversion (and
certainly Max didn't mean it that way), see below for why.
> A veto is supposed to mean, "I disagree with this change. Let's talk
> about it." Stuff is not allowed to ship with vetoed code in it, sure,
> but there should be an assumption of *discussion* rather than
> unilateral action.
>
> I'm not sure how to stress this any more. Let the original committer
> revert the change, or fix it. A vetoer should never do that.
>
> If there is a time pressure due to a release, then the *RM* can
> certainly revert the thing in the release branch (or possibly on trunk
> if that's where the release is coming from; tho I'd recommend a
> branch, a reversion, and then release). The RM is free to take this
> kind of action. Developers should not; this isn't about "easy" or
> "hard" but about social dynamics and the expectation of discussion.
(Are you sure you're not misapplying the RM role from other projects
onto Subversion? Our RM has never really been like that. For
example, any random developer will merge a change into the release
tree, once the change has enough votes. The RM's responsibilities are
mainly about making the release, not about exercising absolute control
over the release tree.)
I think this is just a situation where people may have different
expectations about "the norm". It's like tipping 10% in a country
where 15% is expected: it's not a matter of which number is right or
wrong, it's a question of what conventions are in effect. Along the
border region between two such countries, there might be a bit of
confusion sometimes.
Personally, I'd have no problem with Max reverting a change of mine if
he saw a big problem. Because it's all under version control, I
wouldn't feel that reversion makes the original change any less
accessible -- and therefore, the possibility of restoring it (with a
fix) is no more distant after the reversion than before.
IOW, reversion doesn't mean "I'm banishing this change forever". It
can just mean "This change has a huge problem, let's get it out of the
about-to-be-released tree while the problem is fixed." It's not like
the change has disappeared.
On the other hand, it's not hard to see the reasons behind Greg's and
Julian's position. It makes sense to do the original committer the
courtesy of letting them take responsibility for their own change.
I suggest that in the future, we try to either:
a) Point out the problem, and give the original committer a chance
to fix it (i.e., the Julian/Greg norm), or
b) Revert, but make it clear in the log message that the reversion
is simply because there's a need to get the bug out of the tree
quickly, for whatever reason (in this case, an impending release
candidate), and that the reversion is not intended to banish the
change forever (i.e., apparently the maxb/kfogel norm).
When there's no time pressure, I'd always go for (a) as well. I'm
just saying that when someone does (b), let's not read more into it
than the person intended.
If folks want to formalize a social convention for how
vetos/reversions should be done, go for it. I don't feel a strong
need, though, given that Julian was pretty easygoing about it even
though he was used to a different convention than Max.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Greg Stein <gs...@lyra.org>.
On Wed, Apr 06, 2005 at 02:07:43PM -0500, kfogel@collab.net wrote:
> Julian Foad <ju...@btopenworld.com> writes:
> > Max Bowsher wrote:
> > > On this basis, I am about to revert the revision in question, and
> > > call a vote in the 1.2.x STATUS file for reverting it there too.
> >
> > By the way, it was me that committed that patch on behalf of another
> > contributor, and I would have been happy to revert it myself if you
> > had just pointed out this problem and suggested reverting it. It is
> > usually considered rude to revert someone else's work without first
> > giving them the chance to fix it. Don't worry about it this time,
> > though - I'm cool :-)
>
> Just FYI, I've never felt that way, since it's as easy to unrevert+fix
> the reversion that someone else did. It's all under version control,
> therefore it's all easy :-).
>
> I think Max just wanted to act quickly due to 1.2 scheduling.
I concur with Julian. Reverting somebody else's work is (IMO) bad form.
Recognize that a very likely outcome would be to fix the change rather
than to back it out. By stepping in and reverting the change you are
essentially saying, "I'm vetoing this and there is nothing you can do
about it. There is no other resolution we can come to. Don't even
bother to ask." It completely shuts down any discussion about the
change.
A veto is supposed to mean, "I disagree with this change. Let's talk
about it." Stuff is not allowed to ship with vetoed code in it, sure,
but there should be an assumption of *discussion* rather than
unilateral action.
I'm not sure how to stress this any more. Let the original committer
revert the change, or fix it. A vetoer should never do that.
If there is a time pressure due to a release, then the *RM* can
certainly revert the thing in the release branch (or possibly on trunk
if that's where the release is coming from; tho I'd recommend a
branch, a reversion, and then release). The RM is free to take this
kind of action. Developers should not; this isn't about "easy" or
"hard" but about social dynamics and the expectation of discussion.
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by kf...@collab.net.
Julian Foad <ju...@btopenworld.com> writes:
> Max Bowsher wrote:
> > On this basis, I am about to revert the revision in question, and
> > call a vote in the 1.2.x STATUS file for reverting it there too.
>
> By the way, it was me that committed that patch on behalf of another
> contributor, and I would have been happy to revert it myself if you
> had just pointed out this problem and suggested reverting it. It is
> usually considered rude to revert someone else's work without first
> giving them the chance to fix it. Don't worry about it this time,
> though - I'm cool :-)
Just FYI, I've never felt that way, since it's as easy to unrevert+fix
the reversion that someone else did. It's all under version control,
therefore it's all easy :-).
I think Max just wanted to act quickly due to 1.2 scheduling.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Max Bowsher <ma...@ukf.net>.
Greg Stein wrote:
> On Wed, Apr 06, 2005 at 09:10:16PM +0100, Max Bowsher wrote:
>> ...
>>> Just for my own education as to the process, what is the rationale for
>>> reverting this change as opposed to just allowing it to be fixed?
>>> While I agree that 1.2 should not be released with this change as is,
>>> it seemed like it was something that could have had a fix made for it
>>> relatively quickly and easily. Wouldn't that have been better?
>>
>> I commented that the custom keywords issue had not been fully thought
>> out, and that in resolving it, we might possibly want to extend our
>> svn:keywords syntax. Thus, I propose that we cannot start auto-munging
>> keywords properties until we've designed any addition to the syntax (or
>> decided that we do not need an addition).
>
> You can veto changes on technical grounds. You cannot veto or make
> design decisions, then take unilateral action. That is not right.
>
> If you want to talk about the design, then do so. You can't just start
> reverting stuff because it doesn't match your design choice.
The design was incorrect on technical grounds.
It's not a matter of my choice over someone else's. It's a matter of me
noticing there are important choices involved, which haven't been really
noticed or discussed at all.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Greg Stein <gs...@lyra.org>.
On Wed, Apr 06, 2005 at 09:10:16PM +0100, Max Bowsher wrote:
>...
> >Just for my own education as to the process, what is the rationale for
> >reverting this change as opposed to just allowing it to be fixed? While I
> >agree that 1.2 should not be released with this change as is, it seemed
> >like it was something that could have had a fix made for it relatively
> >quickly and easily. Wouldn't that have been better?
>
> I commented that the custom keywords issue had not been fully thought out,
> and that in resolving it, we might possibly want to extend our svn:keywords
> syntax. Thus, I propose that we cannot start auto-munging keywords
> properties until we've designed any addition to the syntax (or decided that
> we do not need an addition).
You can veto changes on technical grounds. You cannot veto or make
design decisions, then take unilateral action. That is not right.
If you want to talk about the design, then do so. You can't just start
reverting stuff because it doesn't match your design choice.
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Max Bowsher <ma...@ukf.net>.
Mark Phippard wrote:
> Julian Foad <ju...@btopenworld.com> wrote on 04/06/2005 09:53:06 AM:
>
>> Max Bowsher wrote:
>>> On this basis, I am about to revert the revision in question, and call
> a
>>> vote in the 1.2.x STATUS file for reverting it there too.
>>
>
> Just for my own education as to the process, what is the rationale for
> reverting this change as opposed to just allowing it to be fixed? While I
> agree that 1.2 should not be released with this change as is, it seemed
> like it was something that could have had a fix made for it relatively
> quickly and easily. Wouldn't that have been better?
I commented that the custom keywords issue had not been fully thought out,
and that in resolving it, we might possibly want to extend our svn:keywords
syntax. Thus, I propose that we cannot start auto-munging keywords
properties until we've designed any addition to the syntax (or decided that
we do not need an addition).
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Mark Phippard <Ma...@softlanding.com>.
Julian Foad <ju...@btopenworld.com> wrote on 04/06/2005 09:53:06 AM:
> Max Bowsher wrote:
> > On this basis, I am about to revert the revision in question, and call
a
> > vote in the 1.2.x STATUS file for reverting it there too.
>
Just for my own education as to the process, what is the rationale for
reverting this change as opposed to just allowing it to be fixed? While I
agree that 1.2 should not be released with this change as is, it seemed
like it was something that could have had a fix made for it relatively
quickly and easily. Wouldn't that have been better?
Mark
_____________________________________________________________________________
Scanned for SoftLanding Systems, Inc. by IBM Email Security Management Services powered by MessageLabs.
_____________________________________________________________________________
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Julian Foad <ju...@btopenworld.com>.
Max Bowsher wrote:
> On this basis, I am about to revert the revision in question, and call a
> vote in the 1.2.x STATUS file for reverting it there too.
By the way, it was me that committed that patch on behalf of another
contributor, and I would have been happy to revert it myself if you had just
pointed out this problem and suggested reverting it. It is usually considered
rude to revert someone else's work without first giving them the chance to fix
it. Don't worry about it this time, though - I'm cool :-)
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Very emphatic -1 for r13369 (svn:keywords canonicalization)
Posted by Julian Foad <ju...@btopenworld.com>.
Max Bowsher wrote:
> r13369 canonicalizes svn:keywords values.
>
> The canonicalization algorithm is flawed, and silently deletes anything
> that it does not understand.
>
> This means that when we add additional keywords in the future, old
> clients will be unable to edit a svn:keywords property containing a new
> value without inadvertently removing all the keywords which the old
> client version does not support.
Well spotted.
> Also, we have an issue regarding custom keywords. I'm not sure if this
> has been considered yet, but if it is to extend the svn:keywords syntax
> in any way, it is vital that we do not release a version with a
> destructive canonicalization routine.
>
> On this basis, I am about to revert the revision in question, and call a
> vote in the 1.2.x STATUS file for reverting it there too.
>
> Once we have cleared up these issues, we can reintroduce a properly
> designed canonicalization system in 1.3.
+1.
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org