Ruby Forum NGINX > Re: nginx-0.5.36

Posted by Rt Ibmer (Guest)
on 04.05.2008 18:44
(Received via mailing list)
So for the time being should I update the English wiki to show that 
0.6.30 is the latest stable release. It seems this would be best for 
newcomers to understand rather than now it says 0.5.35 is latest.  Yet 
it may also be confusing for some to see 0.6.30 (at least until 0.7 is 
branched) is also the latest dev release.  Any way just let me know what 
you think is best and I'll update it or leave it alone.

The new cache feature in 0.7 sounds interesting - is there any where I 
can learn more about what this is for and its benefits?


      ____________________________________________________________________________________
Be a better friend, newshound, and
know-it-all with Yahoo! Mobile.  Try it now. 
http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
Posted by Cliff Wells (Guest)
on 04.05.2008 19:31
(Received via mailing list)
On Sun, 2008-05-04 at 09:34 -0700, Rt Ibmer wrote:
> So for the time being should I update the English wiki to show that
> 0.6.30 is the latest stable release. It seems this would be best for
> newcomers to understand rather than now it says 0.5.35 is latest.  Yet
> it may also be confusing for some to see 0.6.30 (at least until 0.7 is
> branched) is also the latest dev release.  Any way just let me know
> what you think is best and I'll update it or leave it alone.

Go ahead and mark 0.5.x as "legacy" and 0.6.x as "stable" with a section
mentioning that a 0.7.x "development" branch will soon be started.

Thanks for helping!

Cliff
Posted by Igor Sysoev (Guest)
on 04.05.2008 20:38
(Received via mailing list)
On Sun, May 04, 2008 at 09:34:59AM -0700, Rt Ibmer wrote:

> So for the time being should I update the English wiki to show that 0.6.30 is the latest stable release. It seems this would be best for newcomers to understand rather than now it says 0.5.35 is latest.  Yet it may also be confusing for some to see 0.6.30 (at least until 0.7 is branched) is also the latest dev release.  Any way just let me know what you think is best and I'll update it or leave it alone.

Please, do as Cliff has suggested.

> The new cache feature in 0.7 sounds interesting - is there any where I can learn more about what this is for and its benefits?

In first release cache will cache upstream responses on disk.
Posted by mike (Guest)
on 04.05.2008 23:40
(Received via mailing list)
On 5/4/08, Igor Sysoev <is@rambler-co.ru> wrote:

> > The new cache feature in 0.7 sounds interesting - is there any where I can learn more about what this is for and its benefits?
>
> In first release cache will cache upstream responses on disk.

Is that more or less like this?
http://hostingfu.com/article/nginx-and-mirror-demand
Posted by Igor Sysoev (Guest)
on 05.05.2008 07:29
(Received via mailing list)
On Sun, May 04, 2008 at 02:33:01PM -0700, mike wrote:

> On 5/4/08, Igor Sysoev <is@rambler-co.ru> wrote:
> 
> > > The new cache feature in 0.7 sounds interesting - is there any where I can learn more about what this is for and its benefits?
> >
> > In first release cache will cache upstream responses on disk.
> 
> Is that more or less like this?
> http://hostingfu.com/article/nginx-and-mirror-demand

No, cache takes into account expire time to refresh responses and
inactivity time to delete responses from cache.
Posted by Stefanita Rares Dumitrescu (Guest)
on 05.05.2008 08:43
(Received via mailing list)
sounds like a sweet feature.
Posted by mike (Guest)
on 05.05.2008 10:24
(Received via mailing list)
On 5/4/08, Igor Sysoev <is@rambler-co.ru> wrote:

> > Is that more or less like this?
> > http://hostingfu.com/article/nginx-and-mirror-demand
>
> No, cache takes into account expire time to refresh responses and
> inactivity time to delete responses from cache.

So basically that URL, with an LRU cache mechanism and expiry support?
Posted by Igor Sysoev (Guest)
on 05.05.2008 10:39
(Received via mailing list)
On Mon, May 05, 2008 at 01:11:54AM -0700, mike wrote:

> On 5/4/08, Igor Sysoev <is@rambler-co.ru> wrote:
> 
> > > Is that more or less like this?
> > > http://hostingfu.com/article/nginx-and-mirror-demand
> >
> > No, cache takes into account expire time to refresh responses and
> > inactivity time to delete responses from cache.
> 
> So basically that URL, with an LRU cache mechanism and expiry support?

No, requests are not directly mapped to filesystem. Instead md5 is taken
from the proxied URL and this hash is used to find response in cache.
Besides, the cached response has full header: for example, 404 response
may be cached. Also the response may be cached only if it was requested
no less than, e.g., 2 times for some time.
Posted by Dave Cheney (Guest)
on 05.05.2008 10:50
(Received via mailing list)
Igor - this sounds brilliant !

This is *exactly* what we need in my business to cache a large number
of dynamically generated images that are at the moment being stored in
HUGE monolithic directories.

I look forward to trying this.

Dave
Posted by Igor Sysoev (Guest)
on 05.05.2008 10:55
(Received via mailing list)
On Mon, May 05, 2008 at 06:43:44PM +1000, Dave Cheney wrote:

> Igor - this sounds brilliant !
> 
> This is *exactly* what we need in my business to cache a large number  
> of dynamically generated images that are at the moment being stored in  
> HUGE monolithic directories.
> 
> I look forward to trying this.

I can give the current snapshot.
It ignores the backend "Cache-Control", "Expires", etc. headers, so
this is the main reason why I still did not release it.
Posted by Dave Cheney (Guest)
on 05.05.2008 11:45
(Received via mailing list)
That could be ok, we issue very long ttls on those urls because we
expire them by a version number embedded in the itself.

The main value for would be in hashing the cache store.

Cheers

Dave