Re: [Gnu-arch-users] Ruminations on Arch Desiderata

[Tom Lord]

    > From: Paul Snively <address@hidden>

    > > There is a feature idea that floats around on the back burner for
    > > making it possible to declare a particular file to be of some specific
    > > (non-diff-patchable) type.....

    > That's more what I was inquiring about, not so much the "native" nature 
    > of the AppleSingle file, which is as you say.

See my reply to Doran Moppert.


    >> And, perhaps not.  I try to avoid visiting castles in the sky.

    > That's pretty funny, coming from a guy who's written 
    > [a whole bunch of new software, usually to solve practical
    >  problems that he actually came across].

The "castles in the sky" comment refers to your promotion DNS-SD in a
context where the problem you're offering to solve currently has the
status here of, at most, something that shows up in a fairly
off-the-wall thought experiment.  We're at the stage here of "Ok,
maybe some functionality in that abstract area is something
interesting to think about" and you're at the stage of telling us that
some particular piece of low-level technology is therefore a must
have, that I'm rude for criticising you for that, that I'm
hypocritical because I've written other programs, that I lack a
suitable spirit of competition against some version of CVS that
happens to use that technology .....

What do you expect me to say in reply to something like that?  "Oh,
boy, I'm sold on the idea, let's do it?"

We have a sort of symbolic place-holder in tla for things like
registration services: the `tla grab' command.  It could use things
like a better parser, a formal spec of the language it accepts,
support for configs, careful integration into the transport mechanisms
of src/tla/libarch/archive.h, and some design thought on where to take
it from there (mirroring hints?  branching hints?  patch-queue links?
....).

Now, if you want to layer on top of a fleshed out `grab' mechanism
some kind of contextual resource discovery (why the context should be
defined by network topology is a bit of a mystery, but) -- no doubt
that could then be done entirely externally to arch.

-t