couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Davis <>
Subject Re: svn commit: r1001283 - in /couchdb/trunk/etc/couchdb:
Date Tue, 28 Sep 2010 19:55:32 GMT
On Tue, Sep 28, 2010 at 3:49 PM, Benoit Chesneau <> wrote:
> On Tue, Sep 28, 2010 at 9:41 PM, Paul Davis <> wrote:
>> What if the client doesn't have access to the arbitrarily chosen
>> address? There's no way that CouchDB can guess at all the possible
>> network configurations to try and make that choice. Even if a random
>> address works 99% of the time, why make a decision to break things for
>> the 1%? Even if we list multiple address the client is free to just
>> try the first one and have it work 99% of the time.
> indeed.
> So let' let the client doing the choice itself. I think that's indeed
> case where there will be more than one will be rare. Put all urls on
> different lines may be more "unix".

I'm confused, are you agreeing that we need all the address on the
file system now?

>> I'm not sure what you mean by virtual hosting and different data or
>> what problems that might introduce.
> was thinking to a vhost systeme àla apache
> [vhost]
> port = 5984
> bind_address =
> database_dir = /usr/local/var/lib/couchdb
> view_index_dir = /usr/local/var/lib/couchdb
> uri_file = /usr/local/var/lib/couchdb/couch.uri
> etc . one section per vhost. Just a thought.
> - benoit

Still confused. The basic premise of a vhost is to demultiplex
incoming requests (that share a single transport) and forward them to
separate resources. Ie, forwarding vhosts to different databases. If
each vhost had its own ip:port pair, they wouldn't be vhosts, just
hosts. And further, I don't think I'd be in favor of such a feature as
the technical debt seems a tad high for the benefit.

Paul Davis

View raw message