Here's my take on the tradeoffs. Relative URIs simplify the ALTO server. Assuming the server is built on top of an http handling package, like servlets, the server name and port number are configured at the http handler level. The ALTO-specific server code doesn't see, or otherwise need, that information. The server can extract the server & port from the http handler, but it takes some effort.
On the other hand, relative URIs complicate clients, because the client must resolve the relative URI in the context of the resource directory URI. That isn't hard (most http client libraries have a resolve function), but it is easy to forget. The client will work fine with ALTO servers that return absolute URIs, but will fail the first time it connects to a server that uses relative URIs. It's another example of the Law of Conservation Of Difficulty. :-) - Wendy Roome On 02/12/2013 15:00, "alto-requ...@ietf.org" <alto-requ...@ietf.org> wrote: >Date: Tue, 12 Feb 2013 15:44:48 +0000 >From: "Diego R. Lopez" <di...@tid.es> >Subject: Re: [alto] Should we allow relative URIs in resource > directories? >Hi, > >I think allowing only absolute URIs would translate into simpler >mechanisms to build them, and better understanding by human readers >trying to debug based on ALTO outputs, but I acknowledge these are >limited advantages if there are any other reasons for allowing relative >URIs. _______________________________________________ alto mailing list alto@ietf.org https://www.ietf.org/mailman/listinfo/alto
