General LFS Website affairs

Jeremy Huntwork jhuntwork at linuxfromscratch.org
Mon Oct 18 09:01:26 PDT 2004


On Sun, 2004-10-17 at 23:22 -0400, Anderson Lizardo wrote:
> Yep, actually it's a major rewrite of the procedure we currently use :). By 
> using a makefile, only the things that changed (and others that depend on 
> them) are updated, so an entire website deletion (like we saw a couple of 
> times before) is pratically impossble on this scheme.
> 
> I've re-sent a request for admin help to server-admin (and it's alredy done; 
> thanks Jeremy!). Actually, after the tasks on that e-mail are finished, we 
> will only need to change a SVN property to start updating the actual website 
> (for now it will update http://test.linuxfromscratch.org/rsync-test/; I've 
> also temporarily disabled automatic redirection for TWiki so we can access 
> this page).

Sounds good!

> AFAIK, for now we're still using the old "pull" method, with mirrors receiving 
> a cron time window for mirror-side rsync. One important thing that have 
> stopped our advance with the push method was that we were not using the 
> post-commit method to update the website. This way, even if the push 
> rsync'ing is possible by using a cron job, we would not have the benefit of 
> keeping the mirrors as more up-to-date as possible (the main objective of 
> using push rsync, BTW).
> 
> Now that we have the post-commit running, we can continue making tests 
> (specially evaluating the delay to make a commit) with the push rsync.
> 
> Another thing that got my attention was the not-so-recent rsync flaw regarding 
> rsync on daemon mode. We must make sure that all mirrors that agree to adopt 
> this method have their rsync properly up-to-date regarding this issue.

Definitely.

> I think we have many HTTP mirrors, but fewer FTP mirrors for mirroring LFS 
> packages. We can also suggest those mirrors with plenty space/bandwidth to 
> offer their services for FTP mirroring instead.

Well, something needs to be done in that respect, and we need G's input.

> > 4) Is the repository configured correctly to begin using it in
> > connection with the TWiki? IIRC, we still need a few scripts, right?
> 
> IMHO, the repository is definitely not ready for working with TWiki, at least 
> the way we want (keeping our local modifications on repository and updating 
> the running installation on  commit). The repository reorganization suggested 
> by Jeroen is also missing.
> 
> You just brought up the main TODO items for website (there are some others, 
> like the search engine tuning, putting your lfscounter.cgi script on-line, 
> switching to a CMS, templates for viewcvs, bugzilla, mailman etc.). Right now 
> I think the high-priority items are, in that order: (1) make the post-commit 
> process work, and (2) fix our mirroring scheme, both the method we use and 
> our policy regarding new mirrors. The other items go next.

That seems reasonable. Anything I can do to help with (1) or (2)?

> As you might have noticed, I seem more active on the lfs lists, mainly because 
> I'm starting a reorganization of my schedule to have more time for OS 
> projects :)

Excellent :)

-- 
Jeremy Huntwork
http://www.jenacon.net




More information about the website mailing list