[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kc-kde
Subject:    [kc-kde] Summary #16: SOAP and DCOP
From:       Rob Kaper <cap () capsi ! com>
Date:       2001-07-02 2:17:14
[Download RAW message or body]

Attached.
-- 
Rob Kaper     | 'What? In riddles?' said Gandalf. 'No! For I was talking
cap@capsi.com | aloud to myself. A habit of the old: they choose the wisest
www.capsi.com | person present to speak to; the long explanations needed by
              | the young are wearying.' - "Lord of the Rings", JRR Tolkien


<section
	title="SOAP and DCOP"
	author="Rob Kaper"
	email="cap@capsi.com"
	subject="SOAP <-> DCOP"
	archive="http://lists.kde.org/?l=kde-core-devel&m=99318689017232&w=2"
	posts="30"
	startdate="June 22, 2001 05:15:50 +0000"
	enddate="June 23, 2001 16:52:57 +0000"
> 
<p>
Rik Hemsley asked on kde-core-devel: <quote who="Rik Hemsley">Has anyone
thought of gluing DCOP and SOAP together ?</quote>. Andreas Pour mentioned developers \
could take a look at <a href="http://apps.kde.com/uk/0/info/vid/3344">EasySoap++</a> \
for such an effort, but none of the developers seemed to have worked on a DCOP/SOAP
combination yet, until Dan Pilone posted:</p>
<quote who="Dan Pilone">
<p>
Acutally, there is some work going on.  KDE-Pim work has spawned a few of us
off into basic distributed apps (Mike Pilone, Bryan Dumm, myself).  We're
using SOAP to do the communication between the client (KDE App) and the
server (whatever, but currently either EJBs or Perl).  We have talked about
extending DCOP to do this SOAP communication but haven't acutally done it
yet.  The reason is that SOAP doesn't &quot;quite&quot; line up with what
DCOP is doing now.  Meaning, we could do a lot of what DCOP does, but not
quite all of it.  (The two way registration, the interface discovery, etc.)
We could let the actual calls be carried via soap though.  And as Kurt/Waldo
pointed out, DCOP is currently using libICE as its transport.  We'd need to
put an HTTP transport in there, which would make security an issue again. 
We may end up going the route of kxmlrpcd, where there is a seperate daemon
running that talks to the dcopserver.  We haven't worked that out
yet.</p></quote>
<p>
Apparently Rik Hemsley already did more work than his first post revealed,
because he later mentioned:</p>
<quote who="Rik Hemsley">
<p>
Actually, after I'd got quite far into reading and implementing the SOAP 
specs I realised that SOAP is basically just an extended version of 
XMLRPC. I then looked at your kxmlrpc code and saw it was very similar to 
my own :)</p>
<p>
<i>
Were you thinking of doing a SOAP client library as well?  As in DCOP
to SOAP, not just SOAP to DCOP?</i></p>
<p>
Yes. I've written a class that parses a SOAP envelope and gives you the
parameters you need to make a DCOP call. Right now I can send use a SOAP
message to set my kdesktop background :) Next I'll do the necessary stuff
for doing the reverse.</p>
<p>
I don't know if anyone has a use for this stuff yet, because the SOAP 
spec still needs some work. Perhaps people are using it ? I don't know.</p>
<p>
Anyway, I'll finish the implementation and we can announce it. It will be 
good PR if nothing else :)</p>
<p>
BTW, having read all the docs and written some implementation, I've 
decided that SOAP is a great protocol :)</p>
</quote>
<p>
This caused Martijn Klingens to ask: <quote who="Martijn Klingens">Can it
(technically) replace e.g. DCOP in KDE 3?</quote>. David Faure responded:
<quote who="David Faure">Yeah, sure, replace those short, binary-encoded
messages, with long xml-formatted messages over HTTP. Definitely a win for
performance (and security) :)</quote>. This concern was shared by Kurt
Granoth and research by Christian Gebauer confirmed the performance
issues:</p>
<quote who="Christian Gebauer">
<p>
As the others allready pointed out, that wouldn't be a good idea because of
performance problems. I just did a presentation about SOAP at the university
and according to my research SOAP (and other XML-based protocols) is about
10 times slower than traditional binary RPC/RMI protocols and about 100
times slower than a raw socket connection. This paper compares SOAP with
Java RMI and has a lot of detailed data: <a
href="http://www.extreme.indiana.edu/soap/sc00/paper/index.html">http://www.extreme.indiana.edu/soap/sc00/paper/index.html</a></p>
 </p>
<p>
This is the reason why Microsoft .NET uses a binary protocol for
LANs and SOAP only for the internet.</p>
</quote>
<p>
While it doesn't look like SOAP will replace DCOP, it is good see some
efforts to provide the necessary bindings and bridges that could increase
interoperability between computer systems running different flavours of
software.</p>
</section>



[prev in list] [next in list] [prev in thread] [next in thread]