- From: Martin J. Duerst <mduerst@ifi.unizh.ch>
- Date: Thu, 19 Jun 1997 12:17:19 +0200 (MET DST)
- To: Jim Whitehead <ejw@ics.uci.edu>
- cc: w3c-dist-auth@w3.org
On Wed, 18 Jun 1997, Jim Whitehead wrote: > On June 6, Dylan Barrell wrote: > > 1) The server should support the notion of document language versions > (similar in concept to renditions) and should always produce the language > version corresponding to the language of the user's client, the explicit > language version requested by the client (i.e. it should be possible to > explicitly address a particular language version) or the default language. > 2) The client should be able to specify its preferred languages in order > of preference and the server should present from the existing language > versions the one highest on the client's list. Failing this the default > language should be presented. > 3) It should be possible to specify to the server which language version > is the default (master version) > 4) When asking for a list of objects in a collection it should be possible > to obtain a list of all the different language versions of each resource > (the default should simply return the list of default language resources) > 5) It should be possible to lock only a particular language version of a > resource or all languages > 6) There must be a way of indicating which nodes in the version graph for > each language can be considered equivalent (there might be a different > number of versions for each language version) > > > First, #1, #2, and #3 are already provided by HTTP/1.1 (although there is > currently some debate on this issue). Since we're building on top of > HTTP/1.1, this can be considered base capability. It's true that #1 and #2 are provided. #3 is not really provided. On many servers, it's realized by a (Unix file system) link from document.html to document.en.html, or vice versa. Other conventions exist. I don't know if it's possible to tell the server to make a link using HTTP/1.1, and the mechanisms differ from server to server, so it's not satisfied in a way that would allow to build webdav clients offering this functionality easily. > #4 is interesting, perhaps Content-Language should be returned by INDEX, if > it is defined for a resource. > > As for #5, the view so far has been to assume that every variant of a > resource is itself a resource, and hence has a URI. If it has a URI, then > it can be locked. So this capability is provided for by the locking > proposal. > > #6 can be accomplished using links between resources. Links are discussed > in the properties draft. These are some kind of "semantic links", not links in e.g. an Unix file system? > --- > > To me, the areas within WebDAV which require i18n support are the value > fields for properties and version comments, since these will be displayed > to human operators of WebDAV clients. As a result, I think that the > following requirement could satisfy WebDAV i18n needs: > > Internationalization: All attribute values and version comments must have > provisions for storing one or more of the encoding formats specified in > ISO10646. > > Another way might be to leave it more general, and have a statement like: > > Internationalization: All attribute values and version comments must have > provisions for storing a representation in any human character set. > > Although I'm not sure that ISO10646 and "any human character set" are > identical. I prefer referring to ISO10646 since it's more concrete, and > avoid the issue of which character sets to support. I think for a requirements document, it is enough to address the question of the character repertoire, without reference to encoding issues. So how about: All attribute values and version comments must have provisions for storing all characters from the Universal Character Set (ISO 10646). As for actual implementation, using UTF-8 is probably the best solution. Having several encoding formats complicates things a lot. Regards, Martin.
Received on Thursday, 19 June 1997 06:18:25 UTC