[PEAK] Downloading via redirection

[Phillip J. Eby]

At 12:20 AM 3/24/2007 -0400, Yang wrote:
>On 3/23/07, Phillip J. Eby pje-at-telecommunity.com |python-peak|
><...> wrote:
>>At 08:58 PM 3/23/2007 -0400, Yang wrote:
>> >But when easy_install requests
>> >http://assorted.sf.net/python-commons/download/ (which it apparently
>> >does, based on the output), won't the HTTP response inform it of the
>> >filename and type (and thus it can see that it's downloading
>> >python-commons-0.1.tar.gz)?
>>
>>No, because when it follows a home page or download link, it's looking for
>>an HTML page that it can find download URLs on -- not the actual download
>>file.  At that point, it hasn't chosen *which* URL is the best match for
>>your criteria, so it's not expecting to download anything yet, only to look
>>at web pages for *links* to downloads.
>
>So why not simply check (once you're fetching the download-URL)
>whether it's a web page, as you normally expect (text/html), or
>whether it's in fact a tar.gz (or other acceptable filetype)?

Because:

1. When you're searching for a package by name/version, it needs to know 
the version *before* downloading, so it can pick the best version.  (Which 
includes giving compatible binary distributions precedence over source 
distributions.)

2. It loads web pages into memory, rather than saving them to disk.