Mailing List Archive

On Friday, November 22, 2002, at 06:54 AM, Geoff Richards wrote:

> Hi,
>
> I've been looking into Bricolage for a few days, to evaluate it for a
> client. We've been asked by the IT news website The Register
> (http://www.theregister.co.uk/) whether it would be a good replacement
> for their custom-built CMS. While I've been able to learn enough about
> how it works to put together simpler sites, the Reg has some very
> particular requirements, so I thought it best to run some of them by
> the
> list to see if I'm on the right track.

Cool!

> So, here are the things I'm most concerned about:
>
> * We have in excess of 25000 files published, and we need to be able
> to regenerate everything when, for example, the menus change. Is
> this
> unreasonable? Will bric do it reaseonably efficiently? I've seen
> David suggest SSI, which might be an option, but we do have to avoid
> anything other than prebuilt static pages if possible, because of
> the
> number of hits The Reg gets.

It depends on your definition of "efficient." It also depends on how
complicated your templates are. And it might make a difference which
burner you're using (though probably not -- I think that Bricolage's
API is the slow part, comparatively speaking, not Mason or
HTML::Template). I'm not aware of a system with that many articles in
it, but from what people tell me and from what I've seen, it can take
between 1 and 10 seconds to render a story, depending on how complex
the stories are and how complex the templates are. I think that the
Reg's stories are fairly straight-forward, so I would expect them to
publish very quickly.

> * Can we have multiple 'cover' pages linking to the same set of
> stories?
> We'd need one main index page of recent stories, plus a page for
> each
> section linking to recent stories for that section.

Yes, that's no problem at all.

> * If I publish a new story and the page gets generated, do the
> associated
> cover pages get regenerated to show the new story? If not, is
> there a
> simple way to have this happen automatically?

No, they don't automatically get regenerated. You have to do that
manually. For Macworld, I've been doing this by looking up and
republishing the relevant covers right in the story.mc template, but be
warned that this adds a lot of overhead for a publish. For a site the
size of the Reg, I think I would set up a cron job to run every 5
minutes or so to execute a SOAP job that republishes cover stories so
that they update with all the changes in the last five minutes. The
templates for those covers can even do tricks with MD5 hashes and such
to prevent themselves from being republished if there have been no
changes since the last time they were published.

> * The Reg has load-balanced web servers. If I create an FTP or SFTP
> destination and specify multiple servers, will everything get
> uploaded
> to all of them so that they stay in sync?

Yes, a distribution destination can have as many servers associated
with it as you like. However, doing multiple FTPs will slow down the
publish and distribution process a good deal. A site the size of the
Reg probably ought to have a single file system shared between its
servers (a NetApp would be best). Failing that as a possibility, I
recommend that you have it do a file system copy to a single location
for the distribution, and then set up rsync to update the other servers.

> * We want to build a decent search engine for The Reg (the current one
> isn't great). To do this, we'd want to index the content of each
> page whenever it gets created or updated. Is there a way of
> triggering
> such a thing to happen at publish time? There are these 'actions'
> in the destinations, but I'm not sure if that's what they're for.

Yes, that's *exactly* what they're for. A new, custom action would need
to be developed that knows how to populate the search engine. But since
this is Perl, that should be easily do-able, provided your search
engine has an adequate interface for such things.

> Appreciate any ideas you can offer.

Hope this helps. Let me know if you need any help. I'm happy to consult
on a job like this -- it keeps me involved and allows me to keep making
Bricolage better!

Regards,

David

--
David Wheeler AIM: dwTheory
david@wheeler.net ICQ: 15726394
http://david.wheeler.net/ Yahoo!: dew7e
Jabber: Theory@jabber.org



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Bricolage-General mailing list
Bricolage-General@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bricolage-general

David Wheeler wrote:
> Yes, a distribution destination can have as many servers associated
> with it as you like. However, doing multiple FTPs will slow down the
> publish and distribution process a good deal. A site the size of the
> Reg probably ought to have a single file system shared between its
> servers (a NetApp would be best). Failing that as a possibility, I
> recommend that you have it do a file system copy to a single location
> for the distribution, and then set up rsync to update the other servers.

That actually is exaclty what we do. Works like a charm.

Our sysadmin insists that the site could handle being
slashdotted, and I think he's probably right though we
haven't had that happen yet. I think the Register has it
happen every three or four days right?


--
--
=================================================================
-- mark at geekhive dot net --


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Bricolage-General mailing list
Bricolage-General@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bricolage-general