On 22 Mar 2011, at 21:15, Reinaldo Penno wrote:
> On 3/22/11 1:51 PM, "Ben Niven-Jenkins" <[email protected]> wrote:
>
>> - Obtainment of a service discovery mechanism as URI discovery is a key
>> feature of a RESTful design.
>
> One of the original arguments against URI discovery (as present in v03 of
> the original draft) was P2P clients that are not interested in having a full
> HTTP client built-in can perform a simple GET request and download a file.
> There is no need for the URI discovery phase.
I think you may mean "not interested in having a full ALTO client built-in",
right? URI discovery doesn't affect what HTTP features a client needs to
support but it would be an extra step and would require ALTO clients to first
obtain the "Service Resource" from a well-known URI, process its contents and
then execute a GET to the appropriate URI for the service they are interested
in.
If the burden of implementing logic to process a JSON object containing a list
of URIs (along with their relationship and media type to describe the service
they refer to) as well as logic to process a JOSN object containing an ALTO map
is considered too high, or that needing to do so is likely to cause the primary
target audience to not implement ALTO then avoiding URI discovery would be a
strong motivation.
I personally don't think that the additional burden of URI discovery is very
high at all but I'm not in the P2P client business.
> As far as the 'compromise' approach, I still do not follow the issue of
> 'interoperability'.
Section 10.6 of alto-protocol-07 states:
1. Keep the ALTO Protocol as it is;
2. Restructure this document to allow a REST-ful protocol as an
extension, while keeping the protocol unchanged;
3. Restructure the protocol.
Assuming Option (2) above is the "compromise" approach then in order to build a
RESTful protocol for ALTO it may be necessary to change some of the existing
data types. That may result in effectively forking the protocol into two: one
version using the current data type format and one using the RESTful data type
format. You then have the usual interoperability issues when two versions of
the same protocol exist and may be expected to exist alongside one another.
That might be considered an issue or might be considered "business as usual"
depending on your perspective.
Ben
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
