[PATCH 5 of 5 RFC] RFC: hgwebproxy: an extension for running a caching hgweb proxy

Mads Kiilerich mads at kiilerich.com
Wed Nov 6 13:25:25 CST 2013


On 11/06/2013 07:11 PM, Augie Fackler wrote:
> On Thu, Oct 10, 2013 at 04:35:53AM +0200, Mads Kiilerich wrote:
>> # HG changeset patch
>> # User Mads Kiilerich <madski at unity3d.com>
>> # Date 1381372470 -7200
>> #      Thu Oct 10 04:34:30 2013 +0200
>> # Node ID 7ae5a3498414653b0aa6404b7adfaf3bae63f752
>> # Parent  aaa67a78a1a9e116671082f79342b4f360cd8d2e
>> RFC: hgwebproxy: an extension for running a caching hgweb proxy

This approach turned out to have a number of bugs and design 
limitations. See the other mail announcing 
https://bitbucket.org/Unity-Technologies/hgwebcachingproxy/ .

> Interesting, but I think we should talk about the relative merits of
> this and lookaside clones, since the latter means you could use
> something like AWS to store precomputed most-of-a-repo bundles.

(Your lookaside-bundle repo seems to have disappeared again?)

Agreed - both are attempts at improving scalability ... but in different 
places.

Our concern was not so much the central resources or the leaf resources 
but more the bandwidth to locations with multiple users of the same 
repositories.

It is possible that caching bundles would have advantages compared to 
having a fully expanded repo ... but having a full repo also makes it 
possible to put more logic on the cache server.

/Mads


More information about the Mercurial-devel mailing list