[dais-wg] DAIS Telcon Notes.
Mario Antonioletti
mario at epcc.ed.ac.uk
Wed Jan 19 06:12:01 CST 2005
DAIS Telcon Notes - 18/01/05
=============================
Chair: Norman Paton
Note taker: Mario Antonioletti
Present:
-------
Mario Antonioletti, EPCC
Simon Laws, IBM
Dave Berry, NeSC
Susan Malaika, IBM
Norman Paton, Manchester University
Allen Luniewski, IBM
Agenda:
Framework mapping options for DAIS
DAIS submission planning
Some problem with the number set-up for the telcon. An alternative
number was provided but may have arrived too late for those wishing to
attend the call. Apologies for anyone who wanted to attend this call
but was not able to do so. As we had this problem I am sending this to
the list and will *not* add to grid forge.
---
Susan: Let's talk about the data profiles.
Dave: I attended the EMS call last night. They are trying to sort out
their first profile. I am uncertain as to how an initial data
profile should be structured. I have written emails to Dave
Snelling and Tom Maguire. Dave S. said it still made sense to
talk about DAIS and gridFTP in the same document as you needed to
talk about interoperability of implementations of those
standards.
Norman: if a standard is considered completely in isolation and if
multiple implementations of it are not interoperable then this
is a bug in the standard in which case it might be better to fix
the standard. In WS-I you talk about how you use several standards
together.
Simon: WS-I collects different standards but you don't have to use
them collectively.
Dave: that is my understanding of it ...
Susan: are profiles restricted to GGF standards?
Dave: no
Susan: would the same be true of the data profile? Currently you are
only including GGF standards ....
Dave: you could include other stuff, your CIM work for instance
Susan: there are other things like SQL and XQuery ...
Norman: hope that these would not seep out through the DAIS interfaces
to affect other standards. Whereas if there was a
transactional specification that was agreed on -
WS-Transaction not submitted anywhere and WS-CAF is in OASIS -
then then you may want to be clear about the interaction
between DAIS and WS-CAF.
Susan: could DAIS not hide the transaction?
Norman: up to a point. If you execute a single DAIS request as a
transaction then DAIS could hide that but if someone is trying
to use a WS transactions spec, for example, to group multiple
DAIS messages into a transaction then that could introduce
areas that would have to be clarified ...
Susan: is WS-Transaction in the base profile?
Dave: no
Norman: visible to a service requestor as it would use operations from
the transactional and DAIS specs ...
Dave: I thought DAIS would take whatever query language was
appropriate to the data resource and pass that straight
through....
Norman: that is essentially the case. If you were to use a metadata
description - suppose it used the DMTF metadata description -
then that would align with the SQL99 standard which implies
that any vendor specific metadata would not be described by
this metadata for instance SQL99 would not know the XML
extension of SQL it would not know with what to do with those
extensions ...
Dave: a profile may allow you to use/limit you to use those standards...
Norman: that sounds fair enough. It is tightening something that would
not be covered by the DAIS spec ... would be better though if
this was covered in the DAIS spec but we might miss such things
out ...
Dave: might be genuine competing aims ...
Norman: fair point, the general point though is that in the DAIS specs
there is nothing we can do to clarify the semantic aspects of
transactions specs because these specs do not exist at the
moment and even if they did they could evolve independently in
which case a profile could fill in this gap.
Dave: I would be happy with that . I am trying to clarify within OGSA
the exact role of a profile . There is a lot of pressure in GGF
to produce something , so the aim is to have a basic profile by
summer and an EMS at the end of this year and I suspect they will
want a data profile at the end of the year. The EMS people are
really going for a basic profile - it will not do sophisticated
stuff - just basic job submission attempting to satisfy the
people that have already done that kind of thing.
The latest message from Dave Snelling to OGSA points out that the
OMII are doing most of the basic EMS stuff without using WSRF.
Simon: what prompted that?
Dave: maybe the fact that they are talking to the OMII. They are very
aware of what the OMII are doing .
Norman: if you examine the question as to when DAIS should seek to
standardise - then our position is that we are waiting for
WSRF to stabilise. We also want to ensure that we do not
standardise on things that are not going to be adopted. DAIS
feedback may also change WSRF. We've had a background
supposition that it was inappropriate to standardise without
those things on which we rely already being standards. However
WS-Agreement has gone through with a dependency on WSRF - what
is the OGSA view on this?
Dave: Profiles should only refer to "stable" specification where
"stable" would require presence in a standardisation body, have
multiple interoperable implementations and for the specs to
almost be in its final state. With respect to WS-Agreement not
sure what the politics are going to be.
Norman: in terms of timing for the data profile ... you would like the
DAIS specs to go through the standardisation process even if
WSRF is not through yet as long as the specs are consistent
with the final version of WSRF - right?
Dave: the short answer is "yes" but in order even to start working on
a profile you may be required to have multiple interoperable
implementations ....
Norman: Hursley prototyped the relational spec.
Simon: have prototyped it. It is in the OGSA-DAI CVS - not widely
available ... need to move to get DAIS out but we need to do
that safe in the knowledge that it can be implemented for the
WS-I and the WSRF customers.
Mario: OGSA-DAI will try to implement the DAIS specs.
Norman: yes I realise that, what WSRF implementation did you use?
Simon: we wrote our own.
Dave: as an aside - At present I am still more interested in the
architecture than the profile .
Susan: are the other design teams still doing their work?
Dave: yes. The people doing the profiles may not be the same groups. A
design group does not have to become a profile group.
Susan: is there a deadline?
Dave: still want an OGSA v2 in the summer of this year. Would also
like to get some initial closure. Did you know that there is
going to be another f2f meeting in Feb? Dave Snelling thought it
would be good if someone from DAIS could go ....
Norman: yes, we knew that ... don't know if anyone was thinking of going?
*silence*
Dave: What's happening with the WSRF-DAIS issues?
Susan: people are more aware but ... I think we may see more activity
in the mailing list soon but other than a recognition and some
stuff happening in the mailing list there is nothing else to
report ... we do expect some more items that may be more
relevant ... we need to decide whether we are going to try to
be compatible with WSRF or have a version that is going to be
suitable for folks not interested to WSRF ... have we made a
decision about that yet?
Norman: no, but I have been sending messages to folks about mapping to
WSRF - IBM and Oracle and the data area chairs ... we have not
yet reached a decision. If we are not clear it's difficult to
tell other people ... DAIS must take into account both the
DAIS and WSRF parties. We need to know how big the WSRF
constituency is going to be ... what we do depends on what
answers we get from the major stake holders ... what do the
stakeholders of the DAIS specs think? Whether or not the WSRF
specs will come through in such a way that they minimise the
dichotomy between the WS-I and WSRF communities.
Susan: there have been discussions about putting stuff in the body as
well as the head ... things have not made it to the mailing
list.
Simon: It is valid to have identities in the body that repeat identity
information carried in the header. The messages defined in the
WSRF specifications do not carry such identities.
For WS-I, from an implementation point of view, we'll have
to make something up. Regardless of the mapping we have to
understand the functionality could be used in different
ways. Not possible to have interoperable implementations
without having a mapping ... if we don't put a mapping then if
people implement their own mapping then the likelihood is that
these implementations will not be interoperable ...
Susan: could the profile be used for that?
Dave: that is certainly possible ....
Susan: DAIS would then have to produce the different mappings ...
...
Norman: not very attractive to produce two mappings ... one
universally accepted mapping is better than several ...
Simon: doing more than one mapping is going to be hard work ...
...
Norman: have three main groups of stakeholders:
- IBM
- Oracle
- UK e-Science
it would be good to have feedback from these parties
as well as GGF.
...
If the relevant stake holders get back and agree then we are in
business otherwise we remain in a slightly awkward position.
...
Mario: are we still planning to submit by GGF13? Malcolm keeps on
saying we should submit soon.
Norman: that is what we are trying to decide now ... also Malcolm's
email was what prompted the message to the stake holders.
I have re-read the specs and there are a number of things that
need to be tightened up. There was an action on me to arrange
another call to look at the open issues. We should try to
resolve some of these open issues. So who could work on the
specs?
Simon could do a little, Mario could, Mario said Amy could and Susan
and Norman could do a little. Norman will try to arrange some slots
for spec teams. So we shall try to deal with the issues that are
mapping independent.
Dave: I suspect that you are in a good position to influence WSRF - if
you have the time and effort to make WSRF fit your needs...
Susan: you are suggesting that we highlight the issues with the WSRF
people again?
Dave: yes.
Norman: it has not always been clear what our stake holders views are
and their views have been evolving ... we will try to make
progress with the issues in the specs ... if we were to
produce a spec that depended upon WSRF, does a spec need to be
dependent upon a particular embodiment?
Simon: well ... in theory no. The embodiments are ways of passing
identity ...
Susan: but they would not be interoperable unless they chose a
particular embodiment ... should we refer to WSRF and not
specify an embodiment.
Norman: I was wondering that ... and you do not provide the WSDL...
Simon: but it only affects it from the binding point ...
Susan: difficulties arise though when you request a resource property ...
Simon: WSRF assumes the identity is in the binding ... what it really
does affect is whether our metadata includes references to
anything ... if we use a factory method and return a reference
what does that look like?
Norman: if our stake holders are willing for us to use WSRF then what
do we say in our specs? Could we submit a spec, in the
interim, that did not commit to an embodiment in that
embodiments are not stable yet even if we got a blessing to go
with WSRF ....
Simon: we could do that but we would have to return our own structure ...
Norman: then either everyone commits to one embodiment by the end of
WSRF in which case we join the flow or WSRF uses multiple
embodiments and we use whatever advice is available ...
Dave: that could be done by a profile ...
Norman: yes.
Norman will determine the next telcon date after consulting with Dave
Pearson. A set of telcons for the spec writers will be organised - if
you are interested in joining in this effort then please get in touch.
+-----------------------------------------------------------------------+
|Mario Antonioletti:EPCC,JCMB,The King's Buildings,Edinburgh EH9 3JZ. |
|Tel:0131 650 5141|mario at epcc.ed.ac.uk|http://www.epcc.ed.ac.uk/~mario/ |
+-----------------------------------------------------------------------+
More information about the dais-wg
mailing list