- From: Wilde, Erik <Erik.Wilde@emc.com>
- Date: Mon, 21 Jan 2013 08:49:05 -0500
- To: Roger Menday <roger.menday@uk.fujitsu.com>, Andy Seaborne <andy.seaborne@epimorphics.com>
- CC: "public-ldp-wg@w3.org Group" <public-ldp-wg@w3.org>
hello. On 2013-01-20 21:59 , "Roger Menday" <roger.menday@uk.fujitsu.com> wrote: >>>I just think that what you want to do with LDP, is different from what >>>I want to do. >>> >>> I want solutions where servers which guide my clients through a >>>service. If it is "unfriendly week", the hypermedia directing linking >>>with the :friend predicate shouldn't be there. Your solution is >>>essentially allowing any data to be added. >>> >>> Would you agree ? >> >> I agree that I want LDP to support adding arbitrary data. May not be >> the only usage and some implementations may not allow the opertation >> (thay can always refuse anything). >Well, the server can always refuse something which would otherwise put >the system into an 'illegal' state - but, this isn't the most helpful way >of going about things. i think it would be helpful to always clearly distinguish between protocol data, and payload data. - protocol data is specified in the protocol, interactions must say clearly what kind of protocol data is expected or allowed, and violating any of these constraints violates the protocol and is a client error. - payload data is whatever else is attached to protocol data, and it is of no interest to the server, other than the server accepts and regurgitates it. the server may have some internal restrictions (such as maximum payload size) which mean that there are some limits, but these should just be based on implementation issues. >btw: In the 'graph and links model', a resource might either be liberal >about which other properties it accepts (or not). This then is a model >which covers both arbitrary and constrained data interaction. a LDP server which does not accept non-protocol data would be pretty much useless other than for testing or demo purposes, right? it would simply accept, store and manage the metadata relevant for protocol issues, but would not be able to manage the payload that is the most essential thing for applications. >It seems that most (almost all) applications have some constraints on how >a graph may be evolved by LDP. A concrete, simple example started this >email thread; friends and enemies - if I want to add a new arc to a >Person, called :friend, it is only allowed to link to another person, etc >etc ... that seems like something that's completely out of scope of the LDP protocol. yes, there may be constraints on payload, but defining and enforcing those should now be something LDP is concerned with. cheers, dret.
Received on Monday, 21 January 2013 13:49:51 UTC