[Pgi-wg] PGI TelCon 2009-03-13 Meeting Minutes available for Comments

[Moreno Marzolla]

Andrew Grimshaw wrote:
[...]
> I'm not suggesting that we make WSRF a requirement by any stretch of the
> imagination ... 

I would like to contribute my point of view to the discussion about 
existing specifications. I'm personally not against any existing WS-* 
specification, and I'm willing to implement everything we decide is 
necessary for our "PGI-profile" (or "PGI-specification", or whatever).

However, from a strictly technical point of view, my concerns can be 
summarized by the attached diagram. Each node denotes a WS-* 
specification; the red nodes are the WS-RF specification family. I put 
an arc from X to Y iff in the "Normative References" section of document 
X there is an entry for document Y. Note: this does _NOT_ mean that 
specification X strictly depends on specification Y, but it was easy to 
quickly understand the "relations" among specifications. Furthermore, I 
stopped at the first level (I did not check the "dependencies" of 
WS-Topics, WS-BaseNotification and so forth, but I think that it would 
be a nice exercise to try some day...).

The general point is that the WS-* family of specifications has internal 
dependencies which make implementation efforts quite time consuming, 
because you not only need to support specification X, but also all other 
specifications which X depends upon. Note that there are some libraries 
implementing some of these specifications (in the past, for example, we 
used OpenSAML to support the SAML specification), but nevertheless 
specific hooks to these libraries must be made in the main application 
code, which is still time-consuming.

In my opinion there is a solution around this problem, and is exactly we 
discussed about the WS-SecureAddressing specification. Should we decide 
that a PGI-compliant execution service MUST implement specification 
WS-X, we should explicitly state which (hopefully SMALL) strict subset 
of WS-X MUST be supported; the rest of WS-X of course MAY be supported, 
but that is not required. In this way I hope that implementers won'd 
need to skim through hundreds of pages of additional specifications, and 
needing to implement them as dependencies for our PGI stuff.
For the WS-SecureAddressing stuff we talked about requiring to recognize 
basically a URI embedded inside a small XML fragment containing an 
appropriate namespace telling which security setting the endpoint 
supports. If we are able to state that in a precise and hopefully 
self-contained way, I think that we will make implementers (starting 
from myself :-) ) extremely happy.

Moreno.

-- 
Moreno Marzolla
INFN Sezione di Padova,    via Marzolo 8,   35131 PADOVA,  Italy
EMail: moreno.marzolla at pd.infn.it         Phone: +39 049 8277103
WWW  : http://www.dsi.unive.it/~marzolla  Fax  : +39 049 8756233

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ws-rf.png
Type: image/png
Size: 14858 bytes
Desc: not available
Url : http://www.ogf.org/pipermail/pgi-wg/attachments/20090317/e96f0a53/attachment-0001.png