[Gnu-arch-users] tagline robustness

[Miles Bader]

The tagline tagging method is neat, but I'm a bit concerned about
taglines' sensitivity to seemingly minor changes.

For instance, all the following taglines refer to _different_ tags:

   /* arch-tag: my-gnarly-tag <address@hidden>
    */

   /* arch-tag: my-gnarly-tag  <address@hidden>
    */

   /* arch-tag: my-gnarly-tag <address@hidden> */

   /* arch-tag: my-gnarly-tag <address@hidden>  */

even though to a casual user, they look basically equivalent.

I think this is dangerous, and it offers no appreciable advantage.
Even people that understand (perhaps only peripherally) that arch-tag:
is `special' don't necessarily know all the rules, so it would be better
to make arch try to follow what's natural for humans.

I suggest that should go much further than it does now in reducing
taglines to a canonical form:  instead of just removing preceding and
trailing whitespace, _ignore_ any characters not in a restricted subset.
A good subset might be:  address@hidden

This can be accomplished in a backward-compatible way in two easy steps:

  (1) Change the way arch compares tags to ignore non-subset characters
      in comparisons (and any hash-table indexing or whatever), but keep
      _producing_ tags (e.g. what's written to mod-files-index and
      orig-files-index) using the old rules.

      Change the documentation to say `only these characters are
      significant ...'.

      This way people with old versions of tla/arch/whatever will
      continue to see the same archive contents, and as long as people
      are careful with their taglines (as they've _had_ to be until now,
      so there's not reason for them to change), there's no effect.

      I think it's _extremely_ unlikely that anyone has made taglines
      such that the above change would produce any conflicts.

  (2) A long time from now, maybe change what arch produces (in the
      mod-files-index &c) to include only `significant' characters.

      This step really isn't necessary, of course, but might save a few
      bytes and make things look a bit cleaner.

If there are no objections, I'll look into making a patch that does (1).

Thanks,

-Miles
-- 
We are all lying in the gutter, but some of us are looking at the stars.
-Oscar Wilde