You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by "Gale, David" <Da...@Hypertherm.com> on 2006/02/24 17:20:24 UTC
FW: [DESIGN] Aliases?
robert@infotility.com wrote:
> Quoting "Gale, David" <Da...@Hypertherm.com>:
>
>> Perhaps we could satisfy all of our former CVS-acolytes (myself
>> included) with a "svn alias" command (or some such), which would
>> assign a human-readable name to a specific revision. HEAD, PREV,
>> and the like already exist, but there's currently no way to assign
>> one of the user's own creation. I'd propose something like:
>>
>> svn alias [-force] [-r 21134] NEW_LABEL [path-to-repos]
>>
>> If the command is run within a working copy, the path to the
>> repository should be picked up from that. If no revision is
>> specified, HEAD should be assumed. If the NEW_LABEL alias already
>> exists, an error should be thrown unless -force is specified
>> (matches CVS's behavior). This doesn't change the working copy; it
>> creates a versionless alias on the repository, which would then be
>> available for future "-r <alias>" commands. I'd also suggest a "svn
>> alias -list" command, which would list all of the aliases in the
>> repository and the revisions they point to.
>
> I think all of this is right on the mark.
>
> Since we know already that symbolic aliases are useful
> (HEAD, BASE, etc) why not set up a mechanism to allow
> users to make new ones?
>
> Various correspondents have suggested workarounds or
> hacks of various kinds. Thanks guys, but if there's
> evidence that there is widespread desire for this feature
> (namely that it has been solved multiple times by
> multiple persons) why not just build it into the standard release?
>
> Thanks a lot to everyone who has responded.
> I appreciate your help.
>
> Robert Dodier
Any thoughts from the dev list, before I add this to the issue tracker?
Yes, I'm trying to follow the buddy system. :-)
-David (not subscribed to dev; please CC)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: FW: [DESIGN] Aliases?
Posted by Saulius Grazulis <gr...@ibt.lt>.
On Saturday 25 February 2006 12:24, Max Bowsher wrote:
> I'm sorry, but this most recent discussion doesn't appear to have
> covered any new ground compared with its previous incarnations.
True (alas).
> The proposed design replicates major disadvantages of the CVS model, for
> the sole advantage of shorter command lines;
Wrong, labels are not a disadvantage of CVS model, and even if they were,
adding them to Subversion does not give Subversion any disadvantage.
> * Unversioned: You can't tell who or when 'aliases' were created, nor
> track the history of an alias that undergoes changes.
Wrong, you can easily store a committer and date along with the label:
"VERS-2.2:rev12345:saulius:2006-02-26: etc. etc."
> * Unscoped: It would be far too easy for multiple projects in the same
> repository to clash on their alias naming.
Wrong, it is as easy (or as difficult) as not to clash directory of file names
for the multiple projects.
Not to mention that I find putting multiple (unrelated) projects into one
repository strange and inconvenient. Bu this is again a matter of personal
preferences.
> And, most of all, there is the conceptual weirdness of having two
> totally different UI paradigms for supporting the same goal.
Wrong, there is a conceptual weirdness in misappropriating branch mechanism
and special ad-hoc agreements (like "use 'branch/tags/main' repo layout") to
emulate missing tags.
> Regarding the specific example of merge tracking which seemed to be the
> major justification employed in the users@ thread: This is _entirely_
> possible with properties - indeed, see svnmerge.py in contrib/ for a
> very nice front-end to merging, which does far more than is
> accomplishable with 'aliases'.
The argument "you can already do it by using xyz workaround" does not go.
Taking it to the extreme, you could equally argue thatg version control was
possible long before Subversion, and long before any other VC tool for that
matter:
cp -a project/trunk ChangeLog project/revision-1234
so Subversion should be considered obsolete.
Conceptually, Subversion does the same thing, if you want. It is the
efficiency and convenience that we gain by using Subversion. So naturally
there is a wish to straighten all remaining "rough edges".
Regards,
Saulius
--
Dr. Saulius Grazulis
Institute of Biotechnology
Graiciuno 8
LT-02241 Vilnius
Lietuva (Lithuania)
fax: (+370-5)-2602116
tel.: office: (+370-5)-2602556
mobile: (+370-684)-49802, (+370-614)-36366
Re: FW: [DESIGN] Aliases?
Posted by Saulius Grazulis <gr...@ibt.lt>.
On Saturday 25 February 2006 12:24, Max Bowsher wrote:
> I'm sorry, but this most recent discussion doesn't appear to have
> covered any new ground compared with its previous incarnations.
True (alas).
> The proposed design replicates major disadvantages of the CVS model, for
> the sole advantage of shorter command lines;
Wrong, labels are not a disadvantage of CVS model, and even if they were,
adding them to Subversion does not give Subversion any disadvantage.
> * Unversioned: You can't tell who or when 'aliases' were created, nor
> track the history of an alias that undergoes changes.
Wrong, you can easily store a committer and date along with the label:
"VERS-2.2:rev12345:saulius:2006-02-26: etc. etc."
> * Unscoped: It would be far too easy for multiple projects in the same
> repository to clash on their alias naming.
Wrong, it is as easy (or as difficult) as not to clash directory of file names
for the multiple projects.
Not to mention that I find putting multiple (unrelated) projects into one
repository strange and inconvenient. Bu this is again a matter of personal
preferences.
> And, most of all, there is the conceptual weirdness of having two
> totally different UI paradigms for supporting the same goal.
Wrong, there is a conceptual weirdness in misappropriating branch mechanism
and special ad-hoc agreements (like "use 'branch/tags/main' repo layout") to
emulate missing tags.
> Regarding the specific example of merge tracking which seemed to be the
> major justification employed in the users@ thread: This is _entirely_
> possible with properties - indeed, see svnmerge.py in contrib/ for a
> very nice front-end to merging, which does far more than is
> accomplishable with 'aliases'.
The argument "you can already do it by using xyz workaround" does not go.
Taking it to the extreme, you could equally argue thatg version control was
possible long before Subversion, and long before any other VC tool for that
matter:
cp -a project/trunk ChangeLog project/revision-1234
so Subversion should be considered obsolete.
Conceptually, Subversion does the same thing, if you want. It is the
efficiency and convenience that we gain by using Subversion. So naturally
there is a wish to straighten all remaining "rough edges".
Regards,
Saulius
--
Dr. Saulius Grazulis
Institute of Biotechnology
Graiciuno 8
LT-02241 Vilnius
Lietuva (Lithuania)
fax: (+370-5)-2602116
tel.: office: (+370-5)-2602556
mobile: (+370-684)-49802, (+370-614)-36366
Re: FW: [DESIGN] Aliases?
Posted by Max Bowsher <ma...@ukf.net>.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Gale, David wrote:
> robert@infotility.com wrote:
>> Quoting "Gale, David" <Da...@Hypertherm.com>:
>>
>>> Perhaps we could satisfy all of our former CVS-acolytes (myself
>>> included) with a "svn alias" command (or some such), which would
>>> assign a human-readable name to a specific revision. HEAD, PREV,
>>> and the like already exist, but there's currently no way to assign
>>> one of the user's own creation. I'd propose something like:
>>>
>>> svn alias [-force] [-r 21134] NEW_LABEL [path-to-repos]
>>>
>>> If the command is run within a working copy, the path to the
>>> repository should be picked up from that. If no revision is
>>> specified, HEAD should be assumed. If the NEW_LABEL alias already
>>> exists, an error should be thrown unless -force is specified
>>> (matches CVS's behavior). This doesn't change the working copy; it
>>> creates a versionless alias on the repository, which would then be
>>> available for future "-r <alias>" commands. I'd also suggest a "svn
>>> alias -list" command, which would list all of the aliases in the
>>> repository and the revisions they point to.
>> I think all of this is right on the mark.
>>
>> Since we know already that symbolic aliases are useful
>> (HEAD, BASE, etc) why not set up a mechanism to allow
>> users to make new ones?
>>
>> Various correspondents have suggested workarounds or
>> hacks of various kinds. Thanks guys, but if there's
>> evidence that there is widespread desire for this feature
>> (namely that it has been solved multiple times by
>> multiple persons) why not just build it into the standard release?
>>
>> Thanks a lot to everyone who has responded.
>> I appreciate your help.
>>
>> Robert Dodier
>
> Any thoughts from the dev list, before I add this to the issue tracker?
I'm sorry, but this most recent discussion doesn't appear to have
covered any new ground compared with its previous incarnations.
The proposed design replicates major disadvantages of the CVS model, for
the sole advantage of shorter command lines;
* Unversioned: You can't tell who or when 'aliases' were created, nor
track the history of an alias that undergoes changes.
* Unscoped: It would be far too easy for multiple projects in the same
repository to clash on their alias naming.
And, most of all, there is the conceptual weirdness of having two
totally different UI paradigms for supporting the same goal.
Regarding the specific example of merge tracking which seemed to be the
major justification employed in the users@ thread: This is _entirely_
possible with properties - indeed, see svnmerge.py in contrib/ for a
very nice front-end to merging, which does far more than is
accomplishable with 'aliases'.
Max.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Cygwin)
iD8DBQFEADBbfFNSmcDyxYARAgDWAKCWuqd9lcHHWyo1+PQHIDLIINQ0QgCcCAac
vJLWHEthNXqF+muDyomn3ZM=
=Y0/L
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: FW: [DESIGN] Aliases?
Posted by Max Bowsher <ma...@ukf.net>.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Gale, David wrote:
> robert@infotility.com wrote:
>> Quoting "Gale, David" <Da...@Hypertherm.com>:
>>
>>> Perhaps we could satisfy all of our former CVS-acolytes (myself
>>> included) with a "svn alias" command (or some such), which would
>>> assign a human-readable name to a specific revision. HEAD, PREV,
>>> and the like already exist, but there's currently no way to assign
>>> one of the user's own creation. I'd propose something like:
>>>
>>> svn alias [-force] [-r 21134] NEW_LABEL [path-to-repos]
>>>
>>> If the command is run within a working copy, the path to the
>>> repository should be picked up from that. If no revision is
>>> specified, HEAD should be assumed. If the NEW_LABEL alias already
>>> exists, an error should be thrown unless -force is specified
>>> (matches CVS's behavior). This doesn't change the working copy; it
>>> creates a versionless alias on the repository, which would then be
>>> available for future "-r <alias>" commands. I'd also suggest a "svn
>>> alias -list" command, which would list all of the aliases in the
>>> repository and the revisions they point to.
>> I think all of this is right on the mark.
>>
>> Since we know already that symbolic aliases are useful
>> (HEAD, BASE, etc) why not set up a mechanism to allow
>> users to make new ones?
>>
>> Various correspondents have suggested workarounds or
>> hacks of various kinds. Thanks guys, but if there's
>> evidence that there is widespread desire for this feature
>> (namely that it has been solved multiple times by
>> multiple persons) why not just build it into the standard release?
>>
>> Thanks a lot to everyone who has responded.
>> I appreciate your help.
>>
>> Robert Dodier
>
> Any thoughts from the dev list, before I add this to the issue tracker?
I'm sorry, but this most recent discussion doesn't appear to have
covered any new ground compared with its previous incarnations.
The proposed design replicates major disadvantages of the CVS model, for
the sole advantage of shorter command lines;
* Unversioned: You can't tell who or when 'aliases' were created, nor
track the history of an alias that undergoes changes.
* Unscoped: It would be far too easy for multiple projects in the same
repository to clash on their alias naming.
And, most of all, there is the conceptual weirdness of having two
totally different UI paradigms for supporting the same goal.
Regarding the specific example of merge tracking which seemed to be the
major justification employed in the users@ thread: This is _entirely_
possible with properties - indeed, see svnmerge.py in contrib/ for a
very nice front-end to merging, which does far more than is
accomplishable with 'aliases'.
Max.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Cygwin)
iD8DBQFEADBbfFNSmcDyxYARAgDWAKCWuqd9lcHHWyo1+PQHIDLIINQ0QgCcCAac
vJLWHEthNXqF+muDyomn3ZM=
=Y0/L
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org