11 Apr
2012
11 Apr
'12
5:13 p.m.
Hi Anca,
I hadn't realised this was already in presence_xml. I have added some
comments below.
I would like to make these changes, but it could be a couple of weeks
before I have time. Given that there is a freeze coming up in two weeks
what do you think should be done in the short-term?
On Wed, 2012-04-11 at 16:55 +0300, Anca Vamanu wrote:
The good think about my solution was that I left the
XCAP document
fetch part in presence_xml module. I think it is better like this
because the presence module remains independent of the XCAP part. And
more than this, it works for both integrated and not integrated XCAP
server ( because it uses the generic API of fetching XCAP documents)
and even for any event (not only presence).
I can see that presence_xml is a more logical place to put some of this
implementation.
Here I like better that you have fetched the document
and stored it in
presentity table with expires -1.
As for the fact that you added a new function to be called from the
script when to fetch and store this document (pres_update_presentity),
I think it would have been better to use the existing
pres_refresh_watchers function. If you look in the documentation,
actually when type!= 0 , a pidf manipulation document change might be
expected:
http://git.sip-router.org/cgi-bin/gitweb.cgi?p=sip-router;a=blob;f=modules_…
The reason why it is better to use this function is that it is
actually called by the refereshWatchers MI command and we will
maintain the compatibility with other external XCAP servers that use
this MI command. We can also put a specific type, let's say type=2 for
pidf manipulation document change and type=1 for presentity table
change.
I do agree with this.
When pres_refresh_watchers is called with type=2 , try
to fetch the
pidf manipulation document from xcap server and if found, store it in
presentity table with expires=-1 (as you have done). The only thing
that I would change is to use a wrapper over get_rules_doc function
from presence_xml module. We can do this adding a new field in the
pres_ev_t structure (exactly as get_rules_doc field).
Then continue with the query_db_notify(pres, ev, NULL) function which
will also be called if type=1.
And to maintain compatibility for refereshWatchers function we can
call from here pres_referesh_watchers with type=2, considering that
also a pidf manipulation document
This makes sense to me.
The only thing that I don't know is how to keep
from your
implementation is the possibility for a user to have multiple pidf
documents. I saw that you use the URL to fetch from xcap table and the
file name as ETAG to offer the possibility of storing multiple pidf
documents. However my question is actually whether this use case is
actually possible? Why would a user have more permanent state
documents at the same time?
I did this because I know (from experience) that many different clients
use slightly different file names for the documents they store in XCAP.
This doesn't matter so much for pres-rules, avatars, user-profiles, and
resource-lists because the XML content of the documents is still the
same regardless of name. With presentities it is a bit different. This
is because different clients can (and do) use different XML nodes for
things like <busy/>, <available/>, <vacation/>, as well as supporting
and displaying different fields like emoticons, notes, URLs. This means
that although different clients that support hard-state will both
generate PIDF XML the fields within those XML documents could vary
widely.
Supporting multiple documents for a subscriber at least leaves open the
possibility that someone who uses different clients (one on Windows at
work, one on Linux at home, perhaps) wouldn't be in the situation of
having one client simply dropping stuff the other had PUBLISHed.
For example, the presence client I tested this development with (which
doesn't support hard-state anyway - I had to use curl to upload the
document) didn't support the <vacation/> status from RFC 4827, and
therefore didn't display it (I had to confirm the NOTIFYs were correct
with a Wireshark trace).
Could this ETag stuff be supported by giving an extra (optional)
parameter to pres_refresh_watchers() called ETag?
Peter
--
Peter Dunkley
Technical Director
Crocodile RCS Ltd