Still alive?
Bob Ferris
I just want to hear whether this group is still alive ...
... no not really ;)
However, I recently investigated some time on the topic of change
propagation to close the information flow life cycle. Thereby, I also
updated the related ESW wiki site [1].
I especially added further protocol(s) and vocabulary (hello SPIN ;) )
to the list, because I stumbled about an important issue re. the "last
mile" of change propagation, which is the synchronisation with partly
offline clients (components).
This setting is quite usual in an environment with user agents, or?
Unfortunately, this is exactly the part where the "glorious"
PubSubHubbub protocol fails, because it is intended for server to server
communication and spread updates have to retrieved on time otherwise
they "get lost". So one needs an efficient fallback mechanism that
allows polling.
Since polling on an atomic resource level might quite inefficient, I
would prefer to aggregate such change notification e.g., one can
fragment time periods on different granularities (see for example how
Triplify handle this issue [2]). Although, I can imagine that this might
also be possible with other dimensions (depended domain of the dataset).
All in all, I found for that issue a probably quite applicable solution
from the feed community. It is the Simple Update Protocol [3], which
aggregates digests of changes and clients can poll them to receive the
complete update of subscriptions as needed. Due its different design and
features it can also be combine with PubSubHubbub in the change
propagation chain (see [4]). I especially like the feature using digests
for update notification of resource.
Finally, as already pointed out in the Dataset Dynamics Guide review
thread [5], the resource level is still quite important. That is why I
also like the design of the DSNotify Eventset Vocabulary, where the
triple level is optional. Thereby, I also like the general design of
this vocabulary, because it can be quite good utilized (at least I hope
so ;) ) when performing integrity maintaining tasks in a distributed
knowledge management environment (conflict detection/handling!).
Oh, an issue I almost forgot, which might be quite important: don't
forget to address the push back data to "legacy" sources (existing
non-RDF information services) [6] ;)
Cheers,
Bob
[1] http://www.w3.org/wiki/DatasetDynamics
[2] http://triplify.org/vocabulary/
[3] http://code.google.com/p/simpleupdateprotocol/
[4] http://www.google.com/moderator/#15/e=1a04&t=1a04.40&q=1a04.a1a9
[5]
http://groups.google.com/group/dataset-dynamics/browse_thread/thread/efaed27bb3e0bcc4
[6] http://www.w3.org/wiki/PushBackDataToLegacySources
Michael Hausenblas
Bob,
Not dead, just hibernating. The community seems to have decided that
now the time for implementing and gathering experiences is rather than
trying to standardise or harmonise stuff. Good to see that people keep
maintaining Wiki pages. All input is welcome.
Cheers,
Michael
--
Dr. Michael Hausenblas, Research Fellow
LiDRC - Linked Data Research Centre
DERI - Digital Enterprise Research Institute
NUIG - National University of Ireland, Galway
Ireland, Europe
Tel. +353 91 495730
http://linkeddata.deri.ie/
http://sw-app.org/about.html
Ed Summers
I hadn't seen simpleupdateprotocol before, so thanks for that!
//Ed
Bob Ferris
Hi Ed,
Am 24.02.2011 12:06, schrieb Ed Summers:
> Hi Bob,
>
> I hadn't seen simpleupdateprotocol before, so thanks for that!
Glad, if you like it.
Am 24.02.2011 11:57, schrieb Michael Hausenblas:
>
> Bob,
>
> Not dead, just hibernating. The community seems to have decided that now
> the time for implementing and gathering experiences is rather than
> trying to standardise or harmonise stuff. Good to see that people keep
> maintaining Wiki pages. All input is welcome.
Thanks a lot for the status update, Michael. Generally, it might always
be a good decision to share the knowledge and experience one got while
doing the implementation, or? ;)
(e.g. see Ed's response)
I'm just being interested in hearing your opinion re. my proposed design
decision for implementing change propagation in a distributed knowledge
management system with partly offline components.
Cheers,
Bob