Re: [cross-project-issues-dev] mirror behavior and p2

[David M Williams <david_williams@xxxxxxxxxx>]

If its not obvious, and if it helps clarify
things, this is actually the behavior of the download.php script, and the
"databases" it uses, that return what it knows to be valid mirror
sites. Well, that, and linux file systems. 
This was briefly discussed long ago,
in https://bugs.eclipse.org/bugs/show_bug.cgi?id=238606#c10,
where it was explained to us the the advantages of using a file in a mirrorsURL
instead of a directory. 

Not sure if all that discussion still
applies? Or, if there's anything we can do about it now? ... but, yes,
worth discussing in a new bug. Thus I've opened 

Bug 349945
- Can we improve recommended mirrorsURL value? 






From:      
 Pascal Rapicault <pascal@xxxxxxxxxxxx>
To:      
 Cross project issues
<cross-project-issues-dev@xxxxxxxxxxx>
Date:      
 06/21/2011 08:49 AM
Subject:    
   Re: [cross-project-issues-dev]
mirror behavior and p2
Sent by:    
   cross-project-issues-dev-bounces@xxxxxxxxxxx

---

Hi Stephan, 

I'm a bit surprised by this behaviour. Could you please describe in a bug
reports the steps you took.
Thx

PaScaL

On 2011-06-21, at 4:15 AM, Stephan Herrmann wrote:

> Hi,
> 
> from my latest tests regarding p2.mirrorsURL I observed:
> 
> p2's usage of mirrors seems to depend on the directory timestamp.
> 
> I.e., ANY change to the base directory of a p2 repository will 
> invalidate its mirrors. 
> 
> Since the value of p2.mirrorsURL is this base directory, p2 does not
> consider any mirrors if the directory has changed.
> Such changes would include: editing artifacts.xml to include or 
> modify the p2.mirrorsURL (I was just adding &protocol=http),
> re-adding artifacts.xml to artifacts.jar, re-adding metadata to a
> repository that was kept invisible by omitting those metadata,
> even removing obsolete (backup) files.
> 
> Unless I'm interpreting wrong what I observed, for mirroring to 
> work tomorrow *everything* must be in place today and not be 
> touched in any way. It wasn't obvious to me that this includes 
> touching of the *directory*.
> 
> best,
> Stephan
> _______________________________________________
> cross-project-issues-dev mailing list
> cross-project-issues-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev

_______________________________________________
cross-project-issues-dev mailing list
cross-project-issues-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev