Re: FW: [DESIGN] Aliases?
Tim Hill <[email protected]> Mon, 27 Feb 2006 14:08:01 -0800
| Newsgroups | gmane.comp.version-control.subversion.devel,gmane.mail.eyebrowse.user |
|---|---|
| Organization | Realm Systems |
| Message-ID | <44037841.6080209__12546.0190851756$1141079094$gmane$org@realmsys.com> |
Precisely. I raised this issue last year when pushing for some sort of label scheme, and my motivation was not for CVS aliases. I will re-iterate what I see as the major motivational reason for labels: the reduction in the need to manually enter revision numbers. Here's what I discussed with Julian Ford last year... Essentially, I view an SVN repo as a two-dimensional space of a directory/file tree dimension and a rev# dimension. A file/folder in a repo can thus be located by an (URL,REV#) co-ordinate pair, and this is typically done in many svn commands (the REV# many times being implied). Now, in this model I view a tag as being (partly), a translation of an (URL,REV#) pair directly into an URL. This provides a mnemonic and visible record of this co-ordinate in the repo. However, there is one problem with this translation: tags are not first-class objects within SVN, they are just conventions. There is no command-line way that I can directly express "use the REV# implied by this tag" in SVN. Instead, I have to manually discover the rev# and then manually enter it into an SVN command. This is cumbersome, error-prone and not very scriptable. I assert that THIS is the missing functionality in SVN and, in reality, this is what people really need when they ask for labels: the ability to tell SVN "use the rev# implied by this tag" directly in SVN commands in the same way that HEAD etc are used today. Translation: Tags express rev#s but are only used for OUTPUT. There is an additional need for these rev#s to be available for INPUT into SVN commands without manual transcribing. --Tim Ben Collins-Sussman wrote: >You know, we already have the ability to attach arbitrary unversioned >metadata to revisions. And that's exactly the sort of "label" that >David Gale is talking about. If we had the ability to search >revisions by querying revprops, we'd be done. David could invent all >the revprops he wanted, and search on those revprops to locate >specific revision numbers. > >--------------------------------------------------------------------- >To unsubscribe, e-mail: [email protected] >For additional commands, e-mail: [email protected] > > > > >