[Savannah-hackers] Re: RSS for savannah pages.

[Nic]

Mathieu Roy <address@hidden> writes:

> Can your single RSS be working fine for a project with 30009 bugs,
> 13939943 tasks and 120393 news?

But that's not what RSS is for. RSS is for news. Only recent items
should be included. IMHO we should be including at least one days
bugs, one days tasks and one weeks news. We could switch it round and
just do things by numbers: eg: 5 latest bugs, 5 latest tasks, 5
latest news items.

Note that some versions of RSS limit the number of items that can be
present in the feed to 20. If we were to support these versions using
category based feeds might be a good idea. But I don't think
supporting those versions of RSS is a good idea.


> If the RSS is generated by PHP, only requested RSS would be
> loaded. 

If the RSS were generated on demand then only the RSS that was
requested would cause a resource problem. But RSS is a syndication
tool, it is designed to be aggregated and many aggregators will hit
as many RSS sources as they can find. That's sensible: RSS' value
increases the more points there are in the feed network. But it
creates resource problems for feed providers. 

Ask Slashdot how they feel about RSS.


> And smaller RSS file, no?

Yes. Making 1 file for each category would cause the files to be
smaller. But I don't believe we want the opposite: the fewer the
files the fewer the TCP connections.

I'm not sure how many RSS crawlers use persistent connections (which
alleviates the problem with multiple TCP connections). Curl and WGET
don't maintain persistent connections over requests (because they
can't). Java URLConnections don't use HTTP/1.1 or persistent
connections. I don't know about Perl or Python.


> The tags should, to be coherent with the rest of Savannah (see mail
> headers etc)
> 
> <project>
> <artifact>
> <artifact_id>
> 
> But that's not a problem.

I'm not sure I undestand you... we can't change the RSS tags. That's
how it works: tag conventions.


> Is the RSS, as you seem to be interested in, just about giving last items
> (news, file release -which is more complicated because we do not the
> sourceforge original file release system-, bug)?

RSS is a news syndication tool. You don't use it to syndicate an
entire project description. You could provide a similar XML system to
do that... but you wouldn't want to use the RSS format because very
large files being syndicated would cause a problem.



Your examples:

> The project Tralala use the Savannah bug tracker but
> host his website on www.tralala.org.
> They can use the RSS for the project giving only the bugs to show any
> useful information on www.tralala.org.
> If they have to load a RSS with data for about 1000 projects, it's a
> problem, isn't it?

Why would they need to extract "data for about 1000 projects"? They
just extract the info for their project. In this case what they would
do is have a batch job to do this:

  curl http://savannah-rss.gnu.org/tralala.rss |\
         xslt bug-extracter.xsl - > bugs.html

They can then publish bugs.html on their website. They'd run the
batch a couple of times a day using a curl conditional GET  (-z
cachedfile).



> The project Nipouf use the Savannah task tracker but some people use a
> special software to print nice tabulars with the task tracker
> data... by reading the RSS. Do they need to read a file containing any
> other info?

No. One of 2 things happpens:

- either we provide urls which filter the RSS feeds for each project,
  say:
   http://savannah-rss.gnu.org/nipouf/main.rss
   http://savannah-rss.gnu.org/nipouf/bugs.rss
   http://savannah-rss.gnu.org/nipouf/tasks.rss

  where bugs.rss.xml and tasks.rss are both simply xslted filtered
  versions of main.rss

- or better, the project does it themselves, much like it's done
  above. They provide a new rss file with just the tasks in. Or the
  users get the nipouf project RSS file direct from us and do their
  own xslt job on it (or whatever their RSS reader supports).


> As you see it, what would be the usage of the RSS?

For distribution of news. This is what RSS is for, we should aim to
make it as good as possible at doing that. If, eventually, we can get
to the point where we are updating the rss files for each project
more than 4 times a day we will be doing well.

What do you want it for?


Nic