apt does not handle HTTP redirects
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
apt (Debian) |
Fix Released
|
Unknown
|
|||
apt (Ubuntu) |
Fix Released
|
Wishlist
|
Michael Vogt |
Bug Description
Problem: apt considers 3xx HTTP server responses the
Reproducability: Always
Solution: Make apt's HTTP method follow redirects
Discussion:
I want to set up a package repository on my home (cable modem-based) server that
can scale to many Internet clients.
Since my use is non-commercial, the Coral Cache <http://
be used to distribute the package files. I want to count downloads, so I can
get a sense of how popular my packages are. So, I want to store my package
files in an HTTP-accessible directory and have Apache redirect packages to the
Coralized URL. That way, a client's apt-get install with hit my server, which
will return a "302 Found" redirect to the Coral cache; I count package hits
without serving the huge files.
Unfortunately, apt doesn't handle redirects at all; it treats all 3xx HTTP
responses from the web server as errors.
I have a blog entry here - http://
where I give a tiny bit more detail, but I think this bug report is adequate.
Changed in apt (Debian): | |
status: | New → Fix Released |
description: | updated |
For what it's worth, switching apt's HTTP method to use an existing HTTP library
that is already well-behaved would be one way to solve this problem. One such
library is libcURL.