specification
specification copied to clipboard
Support for MS-Author-Via: SPARQL
Discussed in #85 whether we should use the MS-Author-Via: SPARQL
.
IMHO, redundant because of Accept-Patch: application/sparql-update
.
Moreover, in https://www.w3.org/DesignIssues/ReadWriteLinkedData.html @timbl writes
For a while we advertized the ability to handle patches in SPARQL form like this:
MS-Author-Via: DAV, SPARQL
Now the right way to he ability to handle patches in SPARQL/Update form like this:
Accept-Patch: application/sparql-update
Thus, I think we can safely move this to rough consensus.
+1 to not supporting MS-Author-Via in https://github.com/solid/specification/issues/85#issuecomment-554338829
Moving this to "Done" and closing this issue as there is no requirement to specify.
You don't just switch from one form to another. You always have to have both for a while. Severs in the server-interim must provide both, and clients during the client-interim accept either. Interim periods of a year, for example, would be reasonable. Then when it changes to remove the old one from servers, then you upgrade the major version of the protocol. For example to 1.0.0 in the case of a protocol at 0.9
@timbl : You explicitly OKed this in an Editor's meeting a year ago, and Accept-Patch
was added to NSS six years ago.
I don't think this counts as "just switch". I'm all for smooth transitions, but I would also argue that just arriving at a 0.9 is a major version upgrade from the documents we had earlier.
Tim, point taken but there is also this:
As far as the specs are concerned, there is about a year between ED (created: 2020-12-16) and the Version 0.9.0 (published: 2021-12-17). Or close to two years from the closing of this issue on 2020-02-21 and the publication of Version 0.9.0. Version 0.9.0 was also approved without MS-Author-Via: SPARQL
.
Perhaps we should/could/would've caught potential issues with the tools earlier if we had merged Kjetil's PR back around 2021-10-14 for NSS.
IMO, implementations should transition at their own discretion but Version 0.9.x shouldn't bring backintroduce the header. Please?
If we are going to be consistent about this, then Version 0.9.0 should've kept Accept-Patch: application/sparql-update
as well in addition to current recommendation with text/n3
. I'm sure there are similar cases..