Transport error: Server refuses to fullfil the request

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ben Finney wrote:
| Andrew Bennetts <andrew at canonical.com> writes:
|
|> (You could also try "bzr -Dhttp -Dtransport branch http://..." and
|> looking in the ~/.bzr.log)
|
| Wow, that generates a whole lot of output. It's too large for
| pastebins (it's over 1 MiB of text).
|
| Here is a truncated version, showing only the first few and last few
| transactions.
|

...

There seem to be 2 issues at play here.

#1) We are getting a 403 for a URL that we need data from. I don't know why the
server is blocking this specific request

#2) Our error handling could be clearer. Rather than just saying "refuses" we
could report that we are getting a forbidden (403) error.


| 738.741  * About to connect() to proxy:8118(proxy for people.debian.org)
| 738.745  > GET
http://people.debian.org/%7Emvo/bzr/python-apt/python-apt--debian-sid/.bzr/repository/knits/ad/%2544ist%2555pgrade%2556iew%254eon%2549nteractive.py-20060202114049-0ccb648707ee48e3.kndx
| 738.745  > Connection: Keep-Alive
|> > Accept: */*
|> > User-agent: bzr/1.3.1 (urllib)
|> > Host: people.debian.org
|> > Pragma: no-cache
|> > Cache-control: max-age=0
|
| 738.746  < HTTP/1.1 403 Request for blocked URL
| 738.746  < Content-Length: 10687
| < Content-Type: text/html
| < Cache-Control: no-cache
| < Date: Wed, 14 May 2008 07:08:41 GMT
| < Last-Modified: Wed, 08 Jun 1955 12:00:00 GMT
| < Expires: Sat, 17 Jun 2000 12:00:00 GMT
| < Pragma: no-cache
| < Connection: close
|


There are some other oddities with your proxy. Specifically it seems to give
Connection: close on any readv request. At least, it feels like I see far too
many "About to connect() to..." lines.

I don't really know why all of the other .knit/.kndx files copied fine, while
this one is giving a 403. My best guess is directory permissions on the server.

Specifically, the directory:
bzr/python-apt/python-apt--debian-sid/.bzr/repository/knits/ad

Note the 'ad/' at the end, which differs from the successful knits which were
in, say, 'ef/'.

We split up knit files into multiple directories, to avoid having 1 directory
with 20,000 files.

That, at least, is the best I can come up with for why you would get a 403 after
we have been downloading lots of files successfully.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkgq9NYACgkQJdeBCYSNAAMqPgCfU+qa+RZe1Ja8PxmXgfNV5mBp
7SsAoJm7Dqg9mnLncZM3cUIIO+r40Nb1
=MIJi
-----END PGP SIGNATURE-----