From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============2879380863729787444==" MIME-Version: 1.0 From: =?unknown-8bit?q?R=C3=A9mi?= Denis-Courmont Subject: Re: About Connection between PPP and linux Sockets Date: Fri, 20 Aug 2010 15:11:40 +0200 Message-ID: <6d4cb3c3e5039f1b09ef29d548ef1530@chewa.net> In-Reply-To: <1282304015.23399.210.camel@localhost.localdomain> List-Id: To: ofono@ofono.org --===============2879380863729787444== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Fri, 20 Aug 2010 13:33:35 +0200, Marcel Holtmann wrote: > for a mobile telephony stack, pppd is wrong. Once you understand how 3G > missuses PPP for IP and PDP context setup, you see what I mean ;) For mobile telephony, PPP as a _protocol_ is wrong - sure. That being stated, I don't see how pppd as a _software_ is better or worse than any other PPP implementation. Considering the high bandwidth involved in high HSPA categories, and soon LTE networks, going through user space seems like a bad idea. That's why Phonet GPRS has its own kernel network device driver, even though Phonet can also interact with a TUN device. As such, the PPP-only modems are hopeless. Then why bother rewriting a custom PPP implementation that will be suboptimal in any case? oFono PPP is just as non-performant as pppd in case the user-space AT MUX is used. And it seems worse in case the kernel AT MUX or a raw TTY is used (pppd has kernel path then). So I must be missing something. -- = R=C3=A9mi Denis-Courmont http://www.remlab.net http://fi.linkedin.com/in/remidenis --===============2879380863729787444==--