rest-discuss : Message: The Ambiguous Semantics of PUT: Complete or Incomplete Representations

I know the issue of whether a partial representation can be PUT has come up in at least a couple of threads in this list already, eg "Partial PUT..." and "Bass-ackwards" (and has come up in at least a couple of threads in AtomPP). But I have not seen any definitive resolution of the debate. Worse, after a couple of hours doing my homework on the issue (ie searching the AtomPP mailing list and this one), I have discovered that some participants in the debate seem to have changed their minds on the subject (eg Mark Baker and Joe Gregorio).

Here are the most relevant positions I could find on the subject.

I'd appreciate it if someone could clarify if consensus was ever reached in this list or the AtomPP list or any other "authoritative source".

NOTE: I am NOT seeking people's individual opinions on the subject.[1]

I AM seeking any evidence that the "experts" on REST/AtomPP/HTTP (RFC 2616) have come to agreement on this issue. What would be really useful is clarification on whether there is consensus for each of the set of constraints: Does RFC 2616 allow partial representations to be PUT? Does REST? Does AtomPP?

Obviously, this matters a lot more now that Microsoft has published a draft of its Web3S spec, which enables partial updates via PUT (which Microsoft describes as "merge semantics"). One of the justifications for Web3S is that AtomPP does NOT allow partial updates to be PUT. If this is not true, then it might help close the gap between the two specs.

Thanks.

-- Nick

[1] While I am NOT seeking individual opinions on the topic. I would welcome clarifications/corrections from the people whose positions I referred to in the above URL.

Expand Messages

Author

Sort by Date

I know the issue of whether a partial representation can be PUT has come up in at least a couple of threads in this list already, eg "Partial PUT..." and...

Nick Gall
nick_gall_1117

Jun 20, 2007
9:20 pm

IIRC, on atom-protocol at least the issue was left like this: Servers can do what they want. Clients _cannot_ assume that a PUT does a complete replacement;...

John Panzer
zwyrthrop

Jun 20, 2007
10:35 pm

... That is not the same issue. The question is what the message from the client means. 1.) Must a server obey everything in a client message in order to...

Robert Sayre
robertsayre2000

Jun 20, 2007
11:01 pm

... Messages only have useful semantics if both parties understand them. What I recall is that there was no consensus that Atom servers must choose "omit ==...

John Panzer
zwyrthrop

Jun 21, 2007
1:32 am

... The interface is uniform. All this hand-wringing is a result of people claiming that the meaning of a PUT request depends on the server implementation. The...

Robert Sayre
robertsayre2000

Jun 21, 2007
3:28 am

... I think it actually came to consensus that there was no consensus. Though I could be wrong; there could be no consensus about whether there was consensus. ...

John Panzer
zwyrthrop

Jun 21, 2007
11:01 pm

... Not convincing. Try this: How do I remove a category, if there are three categories listed? ... ... -- Robert Sayre "I would have written a shorter letter,...

Robert Sayre
robertsayre2000

Jun 22, 2007
12:53 am

... You'd PUT the whole representation, replacing the entire entry. This always works. It also works for removing the summary. Servers which support...

John Panzer
zwyrthrop

Jun 22, 2007
2:46 pm

... resource)?...

Mike Dierken
mdierken

Jun 23, 2007
12:45 am

... It doesn't fail, but it's less good than a more specific verb. Also, it makes it difficult to use POST to "add a subordinate resource" to a resource when...

John Panzer
zwyrthrop

Jun 23, 2007
5:30 am

... I agree. However some people here do not agree with this, and have maintained that the server is free to do anything it likes, including retaining the old...

Elliotte Harold
elharo@...

Jun 24, 2007
12:27 am

... I think that the majority of the AtomPub working group simply couldn't figure out how to disallow "bad" server behavior in the spec without also outlawing...

John Panzer
zwyrthrop

Jun 24, 2007
3:56 am

... The HTTP spec makes exactly this point. That's why it doesn't define how a PUT request affects the state of the server. I find it very puzzling that Julian...

Robert Sayre
robertsayre2000

Jun 24, 2007
5:11 am

... I agree that (1) and (2) are true and (2) does not change (1). But way back in the thread <http://tech.groups.yahoo.com/group/rest-discuss/message/9120>, ...

Nick Gall
nick_gall_1117

Jun 24, 2007
12:17 pm

... Yes, you can. That is the understanding that is missing. The messages must have unambiguous semantics in order to build working systems, and ...

Robert Sayre
robertsayre2000

Jun 24, 2007
4:42 pm

... So let me guess this straight: 1. The HTTP spec clearly requires replacement semantics for PUT 2. But the HTTP spec also clearly allows servers to ignore...

Nick Gall
nick_gall_1117

Jun 26, 2007
6:16 am

... Maybe read the thesis again? The simplest way to explain the text is exactly what I've written. Maybe I've gone into too much detail about the...

Robert Sayre
robertsayre2000

Jun 26, 2007
7:47 am

... That's not really true. There aren't any requirements on what servers can do as a result of a PUT request, but there are requirements on what response...

Elliotte Harold
elharo@...

Jun 24, 2007
11:26 pm

... Yes, there are a few requirements given in the definition, but none of them seem to relate to this thread. Success codes for success, Error codes for...

Robert Sayre
robertsayre2000

Jun 25, 2007
12:52 am

... I would expect that on the vast majority of servers you'd just PUT an entry without any categories. What AtomPub agreed on was that the atom-protocol spec...

Brendan Taylor
whateley@...

Jun 22, 2007
3:29 pm

... There wasn't consensus on this point, but I for one continue to argue that HTTP requires that a server that substantially changes a client request not...

Elliotte Harold
elharo@...

Jun 24, 2007
12:05 am

... I don't think you even need to go there with "unset". A PUT request requests that the server set the state of the targetted resource to that represented...

Mark Baker
gonga_thrash

Jun 21, 2007
1:38 am

An XML initiative was started to describe partial XML updates (patches) for XML databases. See http://xmldb-org.sourceforge.net/xupdate/xupdate-wd.html#N56860b...

John D. Heintz
johndheintz

Jun 21, 2007
2:55 am

... Mark, you completely lost me with "if interoperability depends on [consensus] rather than protocol, we're in big trouble." As I understand these admittedly...

Nick Gall
nick_gall_1117

Jun 21, 2007
5:49 am

Nick, I think we have consensus that PUT should be "what RFC2616 says", and you correctly pointed out that the problem is that RCF2616 isn't really very...

Julian Reschke
JulianReschke

Jun 21, 2007
8:15 am

... Fixing the ambiguity problem at its root, RFC 2616, sounds sensible to me. I'll take a look at the relevant mailing lists to see what has been discussed. ...

Nick Gall
nick_gall_1117

Jun 21, 2007
5:52 pm

... Little, unfortunately. ... Absolutely. The first step is getting consensus on the issue, the next step is proposing actual spec text. ... I wouldn't think...

Julian Reschke
JulianReschke

Jun 21, 2007
8:54 pm

Nick, ... "Rough concensus and running code" is for when you're working on a spec. My point is that we've already got a spec, and no matter what we might...

Mark Baker
gonga_thrash

Jun 22, 2007
1:06 am

... But when they look to the spec and it is ambiguous, they will look elsewhere. The repeated threads debating PUT semantics on this list and the AtomPP list...

Nick Gall
nick_gall_1117

Jun 22, 2007
8:33 pm

Yahoo! Groups Tips

Did you know...

Messages

Yahoo! My Yahoo! Mail
	Sign In New User? Sign Up
		Tech - Groups - Help

Yahoo! Groups Tips

Did you know...

Best of Y! Groups

Messages