RE: Batch Annotation creation use case

That is the same thought process we had been having. That is reassuring. Thanks for the response.

From: Robert Sanderson<mailto:azaroth42@gmail.com>
Sent: Wednesday, November 15, 2017 11:54 AM
To: public-openannotation@w3.org<mailto:public-openannotation@w3.org>
Cc: public-openannotation@w3.org<mailto:public-openannotation@w3.org>
Subject: Re: Batch Annotation creation use case


You read correctly that there isn't a batch creation API. The protocol builds upon the LDP specification, which also doesn't have batch processes.

One consideration is that the use of HTTP/2, with persistent asynchronous connections, might mitigate many of the downsides of HTTP/1.1 where forging the connection and waiting for the response before moving to the next operation is no longer a bottleneck. The other approach that has been discussed is the use of websockets for bidirectional communication, however it means implementing the HTTP headers as part of the communication payload ... HTTP over websockets over HTTP is as strange as it sounds.

Hope that helps,

Rob

On Tue, Nov 14, 2017 at 8:58 AM, Patrick Cuba <patrick.cuba@health.slu.edu<mailto:patrick.cuba@health.slu.edu>> wrote:
I have read what documentation and examples I could find for Web Annotation and HTTP and there is a space in which I remain without an answer. Is there a protocol for unContained Annotations to be saved (updated/created) in a batch? With one Annotation, it is clear to return a `Location` header with the URL and similarly when creating a new Container.

In my case an interface creates n new Annotations that may share some relationship but do not belong in a Container together, I have a Web Annotation repository that would like to allow some `POST /batch` scenario with an array of JSON-LD Annotations for creating. The existing behavior returns 201 and a comma separated list of URLs in the `Location` header. Errors break the batch and the first error is returned. Is this unusual? It gets expensive to make 40 calls for a new set, so some batch option feels normal, but I cannot find any examples or guidelines.

In this case, we are building the interface(s) and the Annotation Store, but the Annotation Store is open to the public, so we want to put forth an API that is as congruent as possible with existing standards.

Thanks for your thoughts,

Patrick Cuba, Lead Developer
Walter J. Ong S.J. Center for Digital Humanities
Saint Louis University
+1.314.977.4249<tel:(314)%20977-4249>




--
Rob Sanderson
Semantic Architect
The Getty Trust
Los Angeles, CA 90049