REST, distributed filesystems, and AJAX

[debianfoo]

I have three unrelated questions:

1. Are there plans to provide a REST like interface to the zimbra store?
e.g. for an ical, I may just want to say with a GET: http://www.mydomain.com/zimbra/debianfoo/holiday.ical
or
http://www.mydomain.com/zimbra/debianfoo/tasks/overdue

and then get back some xml or JSON

or use a POST for a more complex request

additionally many of the api's I will want to access will only be REST based, so will the AJAX hooks be developed to accomodate this? e.g. NCBI does have some SOAP access, assuming it is no longer broken :-) ...but their REST like eutils are much easier - and for a wiki/file store I have really liked the XML/HTTP interface provided by the Daisy repository server (java based with lucene/jms) and the rich RBAC mechanisms available - I compared this with the allegro stuff recently, but still came out liking what the daisy guys developed (can elaborate if needed)

...but the lines start to blur with using daisy and zimbra, since why wouldn't I want to serve up my calendars using daisy's RBAC and just treating it as another object in the repository 'bag' - if we're talking workflow, then a bpel file is just another object I grab out of the bag and distribute to my workflow cluster

2. Have you used the zimbra server with any of the exisiting distributed file packages - e.g. google has GFS and the nutch guys replicated with their own version, or MogileFS from livejournal, or OpenAFS

I noticed that you mentioned hardware analogues to some of these approaches elsewhere - e.g. SAN, RAID, etc.

although my email reqs are minimal - my distributed storage needs with complex metadata typings (presumably from mySQL, or a RDF store) are huge

since I see zimbra headed in this direction, I'm curious of your views/plans

3. I saw mention from a Zimbra engineer on the need for a single Ajax framework to compete with the efforts ongoing at MS (can't remember where on the Zimbra site, blog i think). Given this statement, have you evaluated the plethora of existing frameworks (dojo, scriptaculous/prototype, ) and looked at how Zimbra might interoperate with these libraries, or even use them as a base(?).
some semi-useful comparisons/reviews are here:
http://wiki.osafoundation.org/bin/vi.../AjaxLibraries
http://www.ajaxpatterns.org/Ajax_Frameworks
another example 'framework', is orbeon's nice xforms implementation:
http://www.orbeon.com/ops/goto-example/welcome
or what these guys did with the dojo toolkit, for in place editing, etc.:
http://www.turboajax.com/turbodbadmin.html

none of this is meant as criticism, i've been really impressed with what zimbra has done - just curious on where I should be drawing the lines when looking at Zimbra's future dev and place within my app architecture (incidentally, my app will also be open source)

in all fairness, I have not loaded the zimbra source yet either - planned to wait for the next release before hacking - so hope I have not ovelooked any obvious answers to my questions

also hope i did not ramble too much, happy to clarify on any points though.

thanks.

[schemers]

funny you should ask, I was just thinking about this last night Yes, we are definitely planning on offering REST interfaces to access the store (in adddition to our existing SOAP/JSON interface).

I'm working on adding some really simple GET apis to do things like a request a calendar folder as a .ics/.rss/.atom/.xml/.html content, or a contact folder as .csv/.vcard/.xml, etc. Initially this will be for user's to access their own content, and then shortly after it will be expanded to access other user's contet and allow people to publish things like calendars, contacts, etc.

roland

[debianfoo]

great! Thanks roland.

Any comments on items 2 and 3?

2, I can understand since you may not have looked at this before - but if the goal is to also move toward document storage in Zimbra, I assume you will be looking at this (?)

but 3, seems very relevant to the future direction of your AjaxTk - and how the components will be defined - since I assume you want to build a community around developing extensions to this toolkit, and there seems to be no shortage of options for other 'ajax/dhtml toolkits' that are either complimentary or overlapping. In short, I'm just trying to determine how nicely AjaxTk will play with others, in the future. If you are simply way ahead of the other efforts, and can state why, I'm happy with this answer also. But I'm assuming this isn't the case since it seems there are nice components elsewhere, not currently available in the zimbra ajaxtk.

thanks.

[KevinH]

I'll grab 2 and 3.

On 2 we don't currently have plans to use a GFS or MogileFS type storage. We will be adding Red Hat cluster to the Network edition and from our testing this is working pretty well.

With regard to 3 (AjaxTk) before we started we looked at all the tookkits that were out there. Remember this was in late 2003 so many of the links you provided weren't as complete or were just starting off as well. So at the time there was nobody else that had a low level SWT-like widget framework. We didn't want to re-invent the wheel and produce a toolkit but we had no choice. So for the time being we don't plan on switching since our toolkit suits our needs pretty well.

[smies]

I agree with KevinH. The ZimbraTK toolkit is very cool en very wel thought of. Why integrate another library (and be depended on them) when you have your own pretty good working version. It not might be perfect now but I beleave that the team behind it are working hard to improve it. I think they are great and ZimbraTK is on of the best libraries around.