On Tue, 21 Jan 2003, Joshua Slive wrote:
> Date: Tue, 21 Jan 2003 13:00:58 -0500 (Est)
> From: Joshua Slive <[EMAIL PROTECTED]>
> To: "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
> Subject: Re: dist/DATE
Apologies to the mirror maintainers ;
I hate to discuss this on this list,
but I'm afraid there is no other place.
> On Tue, 21 Jan 2003, Henk P. Penning wrote:
> > Not just yet; I only began yesterday.
> OK. Just for your information, the "apache way" usually involves sharing
> code early and often. But since it is your code, you can do whatever you
> want with it.
Just for my information: what would 'sharing' mean here ?
Do I fork over my little prog and that's the last I'll see of it ?
I'm very much in favor of sharing but I would also very much like
to play with this stuff a little longer.
Also, the prog is a little tricky; using async perl IO with selects.
The whole thing runs is tree minutes, mostly waiting for the last
few connections to die. I need to do a little more debugging
and error checking.
> > Isn't the mstat.txt file all you need ?
> But if we are going to integrate something
> directly in our cgi script that selects mirrors, I think it would need to
> be hosted on www.apache.org. Putting an external dependency in there
> would be too dangerous.
The whole idea is dangerous anyway.
-- With a single point of observation, you get a distorted view.
So some Brasil sites seem a little slow; that doesn't mean
that these sites are not useful /in Brasil/.
-- however you implement it, you have to build in safeguards
for failure and exceptions (like the .au site); What if your
database fails? What if your retrieval fails unnoticed;
are all mirrors declared dead ? What if the rsyncd fails ?
Tricky stuff.
-- I suggest you set up your database (add modtime stuff) and
write the stuff to select mirrors. Maybe start by just
/ordering/ the presented mirrors by 'modtime' stuff,
that is, fresh mirrors first.
Retrieve 'mstat.txt' hourly; if it succeeds and passes all
saniny checks, put it in the database. (I could write the
retrieval and sanity checks for you.)
I think this will work just fine. If not I'll be happy to
fork over whatever I got.
> If you choose not to share your code, then we
> will probably need to hack something up ourselves.
I want to stress that I like to share my code; but I'd much rather
share the 'service' for now and keep on hacking at this stuff.
> In any case, thanks for your work on this.
You're welcome.
> Joshua.
Regards.
Henk Penning
Henk P. Penning, Dept of Computer Science, Utrecht University \__/ \
Padualaan 14, P.O. Box 80.089, 3508 TB Utrecht, The Netherlands. \__/
Telephone: +31-30-2534106, fax: 2513791, NIC-handle: HPP1 _/ \__/ \
News.answers http://www.cs.uu.nl/cgi-bin/faqwais \__/ \__/ \__/
