Re: Adding a new OSI layer
- From: Josef Moellers <josef.moellers@xxxxxxxxxxxxxxxxxxx>
- Date: Fri, 16 Mar 2007 13:37:46 +0100
chrb81@xxxxxxxxx wrote:
Hi Josef,
Only put into the kernel what must absolutely belong into the kernel! I
see no reason why the Session Layer should be implemented in kernel space.
But if you do not want to change existing applications (which use
interfaces like socket()...) then it would be necessary to change
something either in the kernel or in a system library.
So it depends on whether you have to consider migration or not...
Not quite: you could easily add another library and then require such applications (or their users) to LD_PRELOAD such a library. Haven't tried that, though.
But it should work, as efence uses that to force an application to use modified malloc/free functions. I guess that the socks package does so, too.
Maybe you'd have to use a few tricks to call the "standard" socket() and friends but that could be solved as well. Maybe you'd really look at socks. Haven't used that, though, too.
Josef
--
These are my personal views and not those of Fujitsu Siemens Computers!
Josef Möllers (Pinguinpfleger bei FSC)
If failure had no penalty success would not be a prize
-- T. Pratchett
.
- Follow-Ups:
- Re: Adding a new OSI layer
- From: chrb81
- Re: Adding a new OSI layer
- References:
- Adding a new OSI layer
- From: chrb81
- Re: Adding a new OSI layer
- From: phil-news-nospam
- Re: Adding a new OSI layer
- From: chrb81
- Re: Adding a new OSI layer
- From: Josef Moellers
- Re: Adding a new OSI layer
- From: chrb81
- Adding a new OSI layer
- Prev by Date: Re: Adding a new OSI layer
- Next by Date: Linux kernel, possible useless continue
- Previous by thread: Re: Adding a new OSI layer
- Next by thread: Re: Adding a new OSI layer
- Index(es):