freecwmp
[Top] [All Lists]

Re: [PATCH] freecwmp: make cwmp version configurable

To: Kaspar Schleiser <kaspar@schleiser.de>
Subject: Re: [PATCH] freecwmp: make cwmp version configurable
From: Luka Perkov <freecwmp@lukaperkov.net>
Date: Wed, 22 Feb 2012 19:52:28 +0100
Authentication-results: mr.google.com; spf=pass (google.com: domain of freecwmp@lukaperkov.net designates 10.213.108.83 as permitted sender) smtp.mail=freecwmp@lukaperkov.net
Cc: freecwmp@linux-mips.org
In-reply-to: <4F451966.9070109@schleiser.de>
Mail-followup-to: Kaspar Schleiser <kaspar@schleiser.de>, freecwmp@linux-mips.org
Original-recipient: rfc822;freecwmp@linux-mips.org
References: <1329917025-15282-1-git-send-email-kaspar@schleiser.de> <20120222152329.GA25852@w500.iskon.local> <4F451966.9070109@schleiser.de>
Sender: freecwmp-bounce@linux-mips.org
User-agent: Mutt/1.5.21 (2010-09-15)
Ok, I have been thinking. We should use 1.0 by default and use only
that. We are not using anything from version 1.2. Once we have that we
can add this easily. I think that is the best way to go for now.

On Wed, Feb 22, 2012 at 05:35:50PM +0100, Kaspar Schleiser wrote:
> >Can you explain why you have put this only in v1.2? This is not v1.2
> >specific.
> Well, I just blindly assumed that this was an 1.0 vs 1.2 issue
> without consulting the specs. ;) I'll gladly remove it from the
> version support patch.
> 
> I just checked, the specs say that *if* the server sends cwmp:ID,
> this id *must* be used in any response to that request. I'll look
> into making the client behave like that, maybe we don't need special
> care for nbbs.

We should do as the specs say. It should be easy to add this check, if
you'll have troubles with it I'll hack it.

> >If you have trouble with your ACS because of this we should introduce a
> >new config option or better yet exclude this code with #ifdef's. And add
> >configuration option for your specific ACS in configure script:
> >
> >/configure --enable-acs=nbbs5
> >
> >Take a look how it is done for HDM, grep for ACS_HDM. You define it at
> >compile time using:
> >
> >/configure --enable-acs=hdm
> >
> >For example we have this in src/http/http.c:
> >
> >94 # ifdef ACS_HDM
> >95         status = zstream_http_configure(http_c.stream, 
> >ZSTREAM_HTTP_COOKIES, 1);
> >96 # elif ACS_MULTI
> >97         status = zstream_http_configure(http_c.stream, 
> >ZSTREAM_HTTP_COOKIES, 3);
> >98 # endif
> I hoped to not need any (non-generic) server specific code

Yes, I also want that. HDM sends on each request cwmp:ID so that is why
I assumed that any ACS is also always sending it.

I have read almost all tr069 specs, but I coded it mostly by reading
captures from the wire ;)

To sum it all up:

 * we'll switch to 1.0 by default
 * handle the case when ACS is not sending cwmp:ID

Kaspar, is that ok with you or I have missed something?

Luka

<Prev in Thread] Current Thread [Next in Thread>