Re: jcr-cmis sandbox
On 2 Dec 2008, at 20:04, Dominique Pfister wrote: -- + rest + ws Just as an observation, I think it's insane having two different protocols for this standard. It sounds like two factions in the standards group that could never agree. -- Torgeir Veimo [EMAIL PROTECTED]
Re: jcr-cmis sandbox
Dominique Pfister wrote: Hi, After having had a first look at the CMIS specification, I decided to start off with the jcr-cmis implementation. I therefore created a jcr-cmis sandbox with the following initial structure: jcr-cmis -- + server + rest + ws I intend to start working on the server/rest subtree (where the REST API binding will reside). Any comment/feedback and - even better :) - any contribution/participation on the the server/ws subtree (containing the Web Services Binding) are more than welcome! I guess it'll be harder to find volunteers for the ws part :-) Two thoughts: 1) At some point of time, we'll have to define a mapping from CMIS to JCR (relatively simple) *and* the other way around (now that's harder). So, how to map identifiers (types, paths), what to do with the CMIS relation objects and so on. Should we start a design document (a text file) for that, or would a Wiki work better? 2) I think that having a separate connector for CMIS in addition to WebDAV should be avoided. We essentially would mint different HTTP URLs for the same thing. So maybe not now, but at a later point of time it would be good if we could merge the new functionality into the existing WebDAV stack. BR, Julian
Re: jcr-cmis sandbox
Also, I don't think we should implement any of the HTTP extensions in the AtomPub binding -- they are neither necessary nor desirable. We should show the TC how to implement it right, not just implement whatever they suggest. very good point! this also puts us into a good position to file issues for the CMIS jira ;) regards, david
Re: jcr-cmis sandbox
Hi Jukka, Most of the organizations on the technical committee of CMIS are already heavily involved at Apache either as contributors or as sponsors and are also on the JCR expert group. If there are existing Apache committers from other projects who'd be interested in working on this, then we could simplify things by opening write access in the Jackrabbit sandbox to all Apache committers. looking at this list I already see Paolo as an outside apache committer. so i think this would be a great idea. regards, david
Re: jcr-cmis sandbox
hi julian, thanks for your comments. ... Since functionally the CMIS specification is a subset of the JCR specification it allows a very simple and straight-forward mapping to a fully compliant JCR repository such as Jackrabbit. ... Yes, the more challenging part is the mapping *from* a JCR repository (how to deal with the information loss). yeah, that's true. it seems that the ideal cmis client (in our case) would be an spi client which then in term would expose jcr again. Defining a mapping will be useful, because it could be re-used to define the relation of CMIS and WebDAV. I think the technically most interesting approach would be to enhance WebDAV to carry the information it currently doesn't have (such as node type information), and then to build CMIS as an extension *into* the Jackrabbit WebDAV layer. absolutely. i think this would be particularly interesting given the fact that a lot of the functionality is already defined in webdav. i wonder if it is possibly to steer the tc into that direction. Another thing the CMIS TC should look into are the various proposals for including support for hierarchies into AtomPub (see, for example, http://www.oracle.com/technology/tech/feeds/spec/draft-divilly-atompub-hierarchy-00.html). It seems to me that this problem is generic enough, and the solution should not be specific to CMIS. good point. do you want me to post this an issue for the tc? ...or do you want to post it yourself. regards, david
Re: jcr-cmis sandbox
David Nuescheler wrote: ... Since functionally the CMIS specification is a subset of the JCR specification it allows a very simple and straight-forward mapping to a fully compliant JCR repository such as Jackrabbit. ... Yes, the more challenging part is the mapping *from* a JCR repository (how to deal with the information loss). Defining a mapping will be useful, because it could be re-used to define the relation of CMIS and WebDAV. I think the technically most interesting approach would be to enhance WebDAV to carry the information it currently doesn't have (such as node type information), and then to build CMIS as an extension *into* the Jackrabbit WebDAV layer. ... Another thing the CMIS TC should look into are the various proposals for including support for hierarchies into AtomPub (see, for example, http://www.oracle.com/technology/tech/feeds/spec/draft-divilly-atompub-hierarchy-00.html). It seems to me that this problem is generic enough, and the solution should not be specific to CMIS. BR, Julian
Re: jcr-cmis sandbox
David, thanks for this. From the peanut gallery (Troy McLure moment: hi Jackrabbits, you might remember me for being an initial mentor of Jackrabbit, and being dragged away since graduation, yet keeping a place in my heart for the project) I have been both interested and quite skeptical about CMIS: maybe it's just me being an old fart, but I love protocols more than I fancy APIs, still I can see how CMIS has a fair deal of shortcomings. Having said that: On Thu, Nov 20, 2008 at 7:17 PM, David Nuescheler [EMAIL PROTECTED] wrote: I am currently working in a technical committee on OASIS defining a document management interoperability specification called CMIS [1]. CMIS shoots for a protocol level interoperability between applications and various repository vendors. The specification is in a very early stage and a lot of things need to be addressed [2], but it has peeked the interest of a number of people at Apache already. Yup, that would include myself - again as an optimist hoping that major CMIS issues might get addressed. Also, you might count on a number of engineers in my organization (disclosure: that'll be www.sourcesense.com) who are very much interested in the proposed standard: we have already been thinking about ways to get our hands dirty, and this sounds like a great opportunity. Since functionally the CMIS specification is a subset of the JCR specification it allows a very simple and straight-forward mapping to a fully compliant JCR repository such as Jackrabbit. Indeed. Not to mention that, as you righteously note, having a CMIS implementation working against the JCR RI should be useful to all those repositories exposing a JCR layer, so I can clearly smell useful stuff. Similar to the existing protocol layers (webdav etc) on top of JCR that are already part of Jackrabbit, I would like to propose that we initiate first tests with an implementation in a sandbox project. I think that there are going to be a lot of benefits from such an implementation. First it will allow any JCR implementation to be CMIS compliant automatically (once the specification is released ;) ) and allow us to find the issues to be fixed in the specification itself and drive it into a good direction. That alone is a noble and worthwhile goal. I do believe that a wire protocol could be really useful to the CM world and, assuming that the TC sees the light and doesn't come up with a white elephant, I can see value in the ASF being the good steward of an implementation providing guidance, critique, feedback and continuous reality check to the spec writers. Let me know what you think. FWIW, I would applaud at that effort and would be following it closely. Happy to help, if help is needed! Thanks, -- Gianugo Rabellino M: +44 779 5364 932 / +39 389 44 26 846 Sourcesense - making sense of Open Source: http://www.sourcesense.com
Re: jcr-cmis sandbox
David Nuescheler schrieb: Hi all, I am currently working in a technical committee on OASIS defining a document management interoperability specification called CMIS [1]. CMIS shoots for a protocol level interoperability between applications and various repository vendors. The specification is in a very early stage and a lot of things need to be addressed [2], but it has peeked the interest of a number of people at Apache already. Since functionally the CMIS specification is a subset of the JCR specification it allows a very simple and straight-forward mapping to a fully compliant JCR repository such as Jackrabbit. Similar to the existing protocol layers (webdav etc) on top of JCR that are already part of Jackrabbit, I would like to propose that we initiate first tests with an implementation in a sandbox project. I think that there are going to be a lot of benefits from such an implementation. First it will allow any JCR implementation to be CMIS compliant automatically (once the specification is released ;) ) and allow us to find the issues to be fixed in the specification itself and drive it into a good direction. But most importantly it will provide an platform for interested parties to collaborate in the open on an implementation. Most of the organizations on the technical committee of CMIS are already heavily involved at Apache either as contributors or as sponsors and are also on the JCR expert group. Let me know what you think. At the moment I cannot help much except saying that I think it's great that you try to explore possibilities and make an effort to make ends meet Cheers Michael regards, david [1] http://lists.oasis-open.org/archives/tc-announce/200810/msg3.html [2] http://roy.gbiv.com/untangled/2008/no-rest-in-cmis
Re: jcr-cmis sandbox
Hi, On Thu, Nov 20, 2008 at 7:17 PM, David Nuescheler [EMAIL PROTECTED] wrote: Similar to the existing protocol layers (webdav etc) on top of JCR that are already part of Jackrabbit, I would like to propose that we initiate first tests with an implementation in a sandbox project. Sounds good. We can get started fast in the sandbox and decide later whether the effort should be promoted to Jackrabbit trunk, turned into a subproject, or branched off somewhere else. But most importantly it will provide an platform for interested parties to collaborate in the open on an implementation. Agreed. Especially if many of the interested people are already familiar with Apache and Jackrabbit. Most of the organizations on the technical committee of CMIS are already heavily involved at Apache either as contributors or as sponsors and are also on the JCR expert group. If there are existing Apache committers from other projects who'd be interested in working on this, then we could simplify things by opening write access in the Jackrabbit sandbox to all Apache committers. BR, Jukka Zitting
Re: jcr-cmis sandbox
Hi all, 2 words about me. working for Sourcesense, committer of Apache POI implementing the Open XML format support, very close to the CM world, in particular to JCR; I have been also working on Alfresco for a couple of years. I am very excited by this thread, since I am one of those hoping that CMIS could be, at least, a battering-ram towards better interoperability among CM vendors. The specification is in a very early stage and a lot of things need to be addressed [2], but it has peeked the interest of a number of people at Apache already. Yes, this is because there is already a missing brick for CMIS in order to be successful, that is a Reference Implementation where to build the Specs. Sounds good. We can get started fast in the sandbox and decide later whether the effort should be promoted to Jackrabbit trunk, turned into a subproject, or branched off somewhere else. If there are existing Apache committers from other projects who'd be interested in working on this, then we could simplify things by opening write access in the Jackrabbit sandbox to all Apache committers. So, here is one! I like the proposal to create a sandbox where we can put stuff that could turn out very good stuff or just junk... Yes: sandbox + committers r/w on it would be the simplest approach. Cheers, -- Paolo Mottadelli: http://www.paolomottadelli.com Sourcesense - making sense of Open Source: http://www.sourcesense.com