Date: Mon 11 Nov 2002 - 15:30:31 GMT
On Mon, 2002-11-04 at 06:48, Paul Sladen wrote:
> > (or, alternative, to have a per-context localhost address which will
> > allow a seamless operation of such software).
> Jacques is currently writing code so that binding to the default address
> within a vserver binds to all the IP addresses assigned to the vserver
> rather than the first one (this hasn't been done before because it is a
> technical pain).
> This will open up the possibility of having a local-loopback in the form of
> `127.1.0.ctx' with modifications to the mangling done above so that we still
> appear to be talking to `127.0.0.1'.
I come accross some programs that do not use the "localhost" name and
the ip address in /etc/hosts, but rather blindly bind and communicate
with its numeric value (127.0.0.1).
When binding to this address, the real address is correctly changed to
whatever the context root is, however, there are a few issues which may
need thinking, for example:
Daemons which check the incoming IP address. This will not be 127.0.0.1
but rather the IP-root (as in named and its controls).
To be usefull, the local loopback should be completely transparent for
the vserver. Bindings should work and report address 127.0.0.1 and
connections to 127.0.0.1 should have client ip address 127.0.0.1 (not
the IP root and *not* the 127.1.0.ctx or 127.1.ctxhi.ctxlo address).
This way daemons which accept connections trough the loopback only will
be happy, while unfriendly (and broken) software which uses 127.0.0.1
directly will have much more chances of working.
I would be very willing to contribute programming and testing time and
resources to this, as I'm currently trying to get a piece of commercial
binary-only cr*p to run within a context.