Re: [Gnu-arch-users] names -> tagline method transition

[Harald Meland]

[Tom Lord]

>     > From: Harald Meland <address@hidden>
>
>     > If multiple file -> id mappings already exist "in the wild" (as I
>     > suspect might be the case for widely-used projects; e.g. gcc), you'd
>     > need some kind of file-id aliasing scheme; 
>
> Or, you could just try to discourage that

What does "you" in the above sentence refer to?  The upstream
(initially not too arch-friendly) maintainer?  Or the person doing the
"archification" of the project?

If the latter, is there anything more one can do than 1) sniff around
to see if someone else already have done an useful arch-gateway of the
project in question (so that all you'll have to do is branch from
that), or, failing such an existing gateway, 2) do the job yourself --
and publish the fact that you have created a gateway?

And, where does the line between "useful" and "useless" existing
gateways go?  The original gateway-creator might have elected to use
"names" tagging, and be quite happy with that choice -- a choice
that's likely to be quite insufficient for some other gateway-craving
user.

Would some ponderings on these issues be considered useful on, e.g.,
the "Tracking a project that doesn't use Arch" wiki page?

> and, when you can't, use names-method changesets to gateway.

If I correctly understand what a "names-method changeset" is, it
wouldn't work for gatewaying between branches in which non-symmetric
renaming have taken place, would it?

>     > I suspect that this is what Bruce Perens et al have been talking
>     > about when they want arch's file-id layer to be driven by some
>     > secondary (local?) mapping.
>
> What are you referring to?

Ooops; I meant Bruce *Stephens*, which i recall as more than once
claiming he has issues with the current implementation of explicit
tags; see e.g. the bottom paragraph of
http://mail.gnu.org/archive/html/gnu-arch-users/2003-10/msg00351.html

(On re-reading it, I'm still not sure that I've understood what Bruce
is talking about; IOW, the suspicion I worded above might be plain
wrong. :-)

>     > Of course, if people could be made to use the first (published)
>     > "arch-converted" version of the project instead of rolling their own,
>     > this might not be a problem.  I, for one, however, don't find that
>     > scenario very likely.
>
> But tagging is an incredibly arbitrary, immaterial choice.  Two
> "forks" that don't agree to cooperate in most ways can agree to tags
> without requiring much of either.

Yes -- assuming that the maintainers of at least one of the forks
knows of the other one, and assuming that their choices for tagging
method isn't incompatible.

> We'll see, I guess.   My hunch is that if 7 different people all try
> to tag GNU sed differently, that the larger community will beat them
> into submission right quick.

Yup, but if the situation could have been avoided in the first place,
then that'd be even better; the greater community needn't spend any
time beating the infidels, and the infidels needn't spend any time to
get their trees in line with the community afterwards.
-- 
Harald