On Fri, May 15, 2009 at 05:27:54PM -0500, Shawn Walker wrote:
> http://cr.opensolaris.org/~swalker/pkg-8724/
depot.txt:
- line 310, 337: "a application/" -> "an application/"
- line 375: legal values for collection_type? Or does anything care?
depot.py:
- line 99,106: no terminating comma here, like on 87.
- line 809: could this list be kept in publisher.Repository? Or maybe in
the repositoryconfig code?
- line 820: no backslash
- line 821, 848: mime type could be kept in p5i.py?
- line 833: rewrap
- line 840: another case where we need to return 404 instead of 500 for
some reason, but can't yet?
- line 844,845: return buf.getvalue()?
- line 850: I'm not sure I see the point of being able to name just a
single package.
- line 865: The problem here is that PM gets passed the filename of the
p5i file the browser downloaded already, rather than a URI?
- line 896: wouldn't '"*" in pfmri' be easier to read?
p5i.py:
- shouldn't client/api.py be modified to use these routines?
- line 95ff: Like depot.py and repositoryconfig.py, this also has a lot
of hardcoded knowledge of the p5i file specifics. It should really all
be kept in one place.
- line 124: Why does "origins" have to be there?
- line 214: trailing blank line
One thing I'm really missing here is the absence of publisher and
repository uuids. I'm okay with that not coming into this wad, but I would
dump the alias attribute entirely, since I don't think it's useful without
uuids. In particular the example server you set up has "opensolaris.org"
as the name, and "os.org" as the alias. I thought the point was that we'd
have uuids as the "real" names, and "alias" be the human readable name
people would see in their "pkg publisher" output, so that we could better
control the confusion of having client configured three publishers
publishing the same bits rather than just repointing "opensolaris.org" to
whatever stream they wanted to be using.
Can we switch to using dashes instead of underscores for text attributes?
A couple of bugs in the test server:
- http://ipkg.sfbay.sun.com:8009/publisher/0/ gives me a file that has no
terminating newline.
- http://ipkg.sfbay.sun.com:8009/p5i/0/SUNWipkg\* only mentions SUNWipkg-um
Where/how is "origins" set? Is this something that should be based off of
proxy_base? Or is it the origins of the depots you're a mirror of?
Perhaps future questions:
- should cfg_cache just be the .p5i file instead of having to translate
between .ini and json formats?
- Where do you see the final configuration store being -- in SMF, or in
cfg_cache? My dns-sd patch puts more stuff into SMF, but I suspect
that's not the way you decided to go here at least for convenience.
But assuming the presence of a Python interface to SMF, would you see
preferring that?
> This does represent somewhat of a flag-day for depot users. Notably, the
> new functionality that has been added will not work correctly until users
> update their cfg_cache. In particular, there two sections entitled
> "[publisher]" and "[repository]" will need to be updated.
There's no way to notice this on startup and rewrite the file? Or is that
what you're saying here will happen:
> The easiest way to do this is to stop and start the depot server (not in
> readonly mode) so that the new sections and attributes will be written to
> the configuration file. A sample of what they might look like is below:
Danek
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
