[arch-dev-public] Putting the process of adding new mirrors on hold

2010-09-30 Thread Roman Kyrylych
Currently we have almost 90 active mirrors in the official mirror list
and new requests for becoming an official mirror are appearing quite often.
However, not all mirrors are good ones (outdated, incomplete, admins
don't respond, etc).
I marked inactive many of them during past months (including about 20
during FrOSCon),
and will continue to do this until all of the mirrors meet the requirements.

22 mirrors that are in the official list did not move to the 2-tier scheme yet,
which means we do not know where do they sync from,
admins did not respond or there is no known email address of the admin.
These will be removed from the list very soon.
If you are using some mirror that was removed from the list recently
and you know how to contact the admin - let me know.

Considering the fact that there is no package signing support yet
I don't see a reason why we should have that many mirrors,
especially when they don't meet the requirements.

Recently I had to defer some requests to become an official mirror
from some private sites. I apologize that this caused a frustration of
their admins
and hope that the reasons are understandable.

To make this fair to everyone I'm thinking about (temporary) putting
the process of adding new mirrors on hold.

-- 
Roman Kyrylych (Роман Кирилич)


Re: [arch-dev-public] Putting the process of adding new mirrors on hold

2010-09-30 Thread Guillaume ALAUX
On 30 September 2010 11:41, Roman Kyrylych roman.kyryl...@gmail.com wrote:
 Currently we have almost 90 active mirrors in the official mirror list
 and new requests for becoming an official mirror are appearing quite often.
 However, not all mirrors are good ones (outdated, incomplete, admins
 don't respond, etc).
 I marked inactive many of them during past months (including about 20
 during FrOSCon),
 and will continue to do this until all of the mirrors meet the requirements.

 22 mirrors that are in the official list did not move to the 2-tier scheme 
 yet,
 which means we do not know where do they sync from,
 admins did not respond or there is no known email address of the admin.
 These will be removed from the list very soon.
 If you are using some mirror that was removed from the list recently
 and you know how to contact the admin - let me know.

 Considering the fact that there is no package signing support yet
 I don't see a reason why we should have that many mirrors,
 especially when they don't meet the requirements.

 Recently I had to defer some requests to become an official mirror
 from some private sites. I apologize that this caused a frustration of
 their admins
 and hope that the reasons are understandable.

 To make this fair to everyone I'm thinking about (temporary) putting
 the process of adding new mirrors on hold.

 --
 Roman Kyrylych (Роман Кирилич)


Hi Roman,

On the mirrors list page [1] I see 56 untiered mirrors. May we
know which are the 22 you mention?
Just say so if you would like some help trying to get in touch with
some mirror admins.

For instance I see these French or Belgian mirrors :
These 2 were removed from the mirrorlist
  -ftp.belnet.be
  -ftp.free.fr

And this one is 3 weeks old and untiered:
  -mir1.archlinux.fr

I could try to mail them in French...

[1] http://www.archlinux.org/mirrors/

--
Guillaume


Re: [arch-dev-public] Putting the process of adding new mirrors on hold

2010-09-30 Thread Dan McGee
On Thu, Sep 30, 2010 at 6:56 AM, Guillaume ALAUX guilla...@alaux.net wrote:
 On 30 September 2010 11:41, Roman Kyrylych roman.kyryl...@gmail.com wrote:
 Currently we have almost 90 active mirrors in the official mirror list
 and new requests for becoming an official mirror are appearing quite often.
 However, not all mirrors are good ones (outdated, incomplete, admins
 don't respond, etc).
 I marked inactive many of them during past months (including about 20
 during FrOSCon),
 and will continue to do this until all of the mirrors meet the requirements.

 22 mirrors that are in the official list did not move to the 2-tier scheme 
 yet,
 which means we do not know where do they sync from,
 admins did not respond or there is no known email address of the admin.
 These will be removed from the list very soon.

If they are syncing and up-to-date
(http://www.archlinux.org/mirrors/status/), then why remove them? That
seems pretty silly to me. We one remaining that seems to have fallen
off the map (unix.pl), 5 that are out of date, and one without a
lastsync file.

 If you are using some mirror that was removed from the list recently
 and you know how to contact the admin - let me know.

 Considering the fact that there is no package signing support yet
 I don't see a reason why we should have that many mirrors,
 especially when they don't meet the requirements.

 Recently I had to defer some requests to become an official mirror
 from some private sites. I apologize that this caused a frustration of
 their admins
 and hope that the reasons are understandable.

 To make this fair to everyone I'm thinking about (temporary) putting
 the process of adding new mirrors on hold.

 --
 Roman Kyrylych (Роман Кирилич)


 Hi Roman,

 On the mirrors list page [1] I see 56 untiered mirrors. May we
 know which are the 22 you mention?
 Just say so if you would like some help trying to get in touch with
 some mirror admins.

You probably missed the active column there.

 For instance I see these French or Belgian mirrors :
 These 2 were removed from the mirrorlist
  -ftp.belnet.be
  -ftp.free.fr

 And this one is 3 weeks old and untiered:
  -mir1.archlinux.fr

 I could try to mail them in French...

No one will ever stand in your way of helping out. I'm sure you can
chip in; if you start doing a lot you can probably get in the coveted
Mirror Admin group. :P

-Dan


[arch-dev-public] [signoff] logrotate-3.7.9-1

2010-09-30 Thread Allan McRae

Upstream update.

Signoff both,
Allan.


Re: [arch-dev-public] Putting the process of adding new mirrors on hold

2010-09-30 Thread Roman Kyrylych
On Thu, Sep 30, 2010 at 15:16, Dan McGee dpmc...@gmail.com wrote:
 If they are syncing and up-to-date
 (http://www.archlinux.org/mirrors/status/), then why remove them? That
 seems pretty silly to me.

No known admin email address or no response.
No known upstream.
I agree that these may not be very solid reasons,
but I see no other way to bring things to order.

If we will need to change something within our mirroring scheme
(e.g. use of mirrorbrain or geodns, or some other change)
- we will have the problem contacting mirror admins again.

If we would not disable rsync access for non-tier1 mirrors
on rsync.archlinux.org we would never hear from some mirror admins,
that I was unable to contact until their mirror stopped working
(and BTW, some didn't even notice that at all).

It may be that my frustration influences my view,
so if you think I'm totally wrong - I will keep things as they are now.

-- 
Roman Kyrylych (Роман Кирилич)


Re: [arch-dev-public] Putting the process of adding new mirrors on hold

2010-09-30 Thread Dan McGee
On Thu, Sep 30, 2010 at 11:41 AM, Roman Kyrylych
roman.kyryl...@gmail.com wrote:
 On Thu, Sep 30, 2010 at 15:16, Dan McGee dpmc...@gmail.com wrote:
 If they are syncing and up-to-date
 (http://www.archlinux.org/mirrors/status/), then why remove them? That
 seems pretty silly to me.

 No known admin email address or no response.
 No known upstream.
 I agree that these may not be very solid reasons,
 but I see no other way to bring things to order.

 If we will need to change something within our mirroring scheme
 (e.g. use of mirrorbrain or geodns, or some other change)
 - we will have the problem contacting mirror admins again.

 If we would not disable rsync access for non-tier1 mirrors
 on rsync.archlinux.org we would never hear from some mirror admins,
 that I was unable to contact until their mirror stopped working
 (and BTW, some didn't even notice that at all).

 It may be that my frustration influences my view,
 so if you think I'm totally wrong - I will keep things as they are now.

First, I 100% understand the frustration.

I'm only saying keep it in perspective- if we end up making a change
that breaks these mirrors that have gone MIA, that is fine, and
indicates forward progress on some other front. But breaking them on
purpose just seems silly, if there is no need (read: forward progress)
to.

If we ever do mirrorbrain or geodns, then we can make the necessary
changes at that time rather than doing anything ow when (as far as I'm
aware) nothing big is on the horizon, or needs to be.

I'd propose leaving these untended mirrors in the list, but *only* if
they magically stay up to date and relatively pain-free. If they begin
to lag, then mark them as inactive and they will no longer be on the
official list.

-Dan


Re: [arch-dev-public] Mirror status page

2010-09-30 Thread Dan McGee
Suggestions and things are quoted below.

On Tue, Sep 21, 2010 at 8:41 PM, Sander Jansen s.jan...@gmail.com wrote:
 Hi Dan,

 Just nitpicking here: putting Unknown into the columns messes up
 sorting for certain columns. For example when sorting the out-of-sync
 mirrors by score, from best-to-worst, it will put the unknowns on
 top e.g best. It would make more sense to put them on the bottom e.g.
 worst.

Thanks! Fixed locally and will get pushed out.

On Sat, Sep 25, 2010 at 5:31 AM, Pierre Schmitz pie...@archlinux.de wrote:
 On Thu, 23 Sep 2010 09:52:55 -0500, Dan McGee dpmc...@gmail.com
 wrote:
 On Thu, Sep 23, 2010 at 1:53 AM, Pierre Schmitz pie...@archlinux.de wrote:
 * You should rethink about putting those equations into yellow boxes
 with a bigger font size. They are not really the most important content
 on that site.
 This really has more to do with the theme/styling on the entire side
 and less to do with this page. I honestly don't think it distracts too
 much, but if a lot of people do we should adjust the theme. It is odd
 that the tt/ and code/ tags don't really match up; you can see I
 used both of them.

 Should I just switch to tt/ everywhere?

 No idea. But I'd reduce the font size to match the one from the
 surrounding text. It's not distracting, but clearly highlighted. But I
 don't care that much about this...

tt/ everywhere it is.

 * One should add to the μ Delay that it has an error of about an hour
 (=frequency of your cron job). Especially those with a delay of less
 than an hour sync hourly for real.
 I'm not sure I follow your logic- if they sync on the top of the hour,
 then we're pretty close to the average delay, which is halfway
 between 0-59 minutes anyway. I think it might be worth scheduling the
 cronjob in a slightly different way to not give any mirror an
 advantage. If we sync on the following schedule (13 times in 12
 hours), no one time would get the benefit:
 0:04 0:59 1:54 2:49 3:45 4:40 5:35 6:30 7:26 8:21 9:16 10:11 11:07

 Sounds confusing of course, but would this make things better?

 Should work. Another option would beto just round up the values to
 hours.

This should all be taken care of.

 * Make the mirror urls clickable
 I thought of this, but not sure what to click through to. We can
 easily do a mirror info page, but we can't/shouldn't show everything
 publicly.

 I was more thinking about just putting an a tag around the url. :-)

Seems a bit silly to link to the mirror itself only because there
isn't much of use there. If you want a package, we have the download
package link, etc. But there are some more changes coming that expose
the mirror objects a bit more publicly via /mirrors/.

 * Would be nice if you could use the iso country names. E.g. its not
 Great Britain but United Kingdom or Russian Federation instead of
 Russia.
 We've been using these same names for ages. I can give you permission
 to go in and update things as you see fit if you'd like- the country
 is just freeform right now, there aren't choices.

 So I could just update these values in the db without any side effects?

 Getting this tied to an actual country object could make GeoIP
 integration of some sort be useful later down the road.

 That was my intention.

Haven't seen any progress on this. :)

I'm going to push the remaining changes public soon, and then publish
a news item if that is OK highlighting the new page. If there are any
more objections or feedback please get it in now.

-Dan