Per Einar Ellefsen wrote:Link to perl.apache.org/dist/, like we're doing right now. Mirrors won't even have to mirror /dist (it isn't in CVS).
"won't have it" is easy to say, in practice they will, even if you use robot.txt
Oh yeah, you think of mirroring as using some package that follows links to do it. I was thinking of using CVS and bin/build.
but it's not only /dist, it's also the info about the latest releases.
also if there are bugs in docs we fix them, users using outdated mirrors won't get these fixes.
Search will also won't work on mirrors unless mirrors will take the pain to setup special environment, which I doubt they will. This will only frustrate people and they will end up at the master site.
Point the search to perl.apache.org.
think further, once they did the search they end up on the master site, how can they go back? Will mirrors set a smart rewrite? I doubt so, will we do such a thing, I doubt too.
Yeah, true, but in that case we would atleast have the benefit of mirrors for pages that aren't searched. For example, I have my local copy of the site, but I haven't set up the search. For that I just go to perl.apache.org and browse on from there.
Probably there are other reasons for not supporting any official mirrors.
I suggest that we simply let people do what they want and have nothing official (which they will without asking us). We can link to the sites that mirror, without any guarantees what so ever. In this way we don't have to worry about outdated or broken mirrors, and keep our minds concentrated on other more useful things.
Well, if we have something semi-official, at least it does give us better control. We don't have to have tons of different mirrors, but most likely mirrors that we trust. For example, Randy would maybe set up a mirror (like he did with the guide): we can probably trust him to take the correct measures. And by the way: the mod_perl site is perl.apache.org. Basta. People will realise that.
Well, why do you let Randy or Thomas mirror it (and link to his site) but not to Joe fat-bandwidth? Why inventing troubles?
Because of the reasons mentioned: mirroring implies a trust-network. Why invent troubles? because we would probably get more benefit than loss out of mirroring. You have to balance things, and not just see the drawbacks here. Not everybody can browse the site as quickly as you can, and I am usually very happy when I see a site that links to a mirror in Europe: that's how we could go about it: have a mirror in each continent that we know are up to date because we either know or manage the mirror. Like that, there aren't too many to watch out for, and we can still have the benefit of the mirrors.
That's why the best solution I see is having nothing official and have a page with links to other sites that may or may not be up-to-date, complete, working or what not.
also remember (which you probably don't) that the guide was usually released once in a few months (every months in the fast growth days), so it was easy to update mirrors. Now we move to the new mode, where updates are immediate (well 6 hours between updates, with manual immediate update on demand), so mirroring has to be more pro-active.
Yes, I know. But I actually think that our current situation makes it *easier* to be up-to-date, because updates can be done by cron, while before they had to be downloaded manually.
-- Per Einar Ellefsen [EMAIL PROTECTED]
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
