RE: [hpoj-devel] patch to improve scan speed/robustness

[PASCHAL,DAVID (HP-Roseville,ex1) <dav...@hp...>]

Hi, Gerhard.

> can you remember the problem with MLCConfigSocket?
> If a 2nd connection to the same peripheral port was attempted, then
> ieee12844.c sent a 2nd MLCConfigSocket although this violates the
> protocol. Therefore the 2nd connection to the same peripheral port
> always failed.
> The driver should remember to which ports it already has sent
> MLCConfigSocket and also the negotiated parameters, and should not
> send it at the 2nd connect attempt.
> Do you remeber: Some time ago you told me about the problem, 
> and I told
> you, that I know, but that I hadn't it implemented yet.
> Do you have already fixed this problem?
I haven't fixed that problem, and I had sort of forgotten about it until
now.  The driver is probably locking up in response to the Error packet from
the peripheral.

Actually, I have made few or no changes to ieee12844.c since I took over.
The vast majority of my kernel-mode hacking has been on ieee12844pp.c to
improve the parallel signaling (such as the patch that started this thread).
I probably won't fix things like the duplicate ConfigSocket problem unless
absolutely necessary, because eventually I plan to replace the kernel-mode
drivers with a user-mode daemon anyway (and add support for true 1284.4,
USB, PML multiplexing, and non-Linux platforms).

> But even if this problem is fixed, then I'm not sure whether the
> OfficeJet will a) accept a 2nd simultaneous PML connection and b)
> whether the OJ will correclty multiplex the PML 
> requests/responses/traps
> over the two connections. According to the PML spec one could assume,
> that PML devices _are_ generally capable to do this. But I wouldn't
> believe it before I've seen that it works ...
> If it works (with all OJ models), then maybe you can omit a 
> separate PML
> multiplexor on the host, if not, then you will definitely need it.
Most HP peripherals don't support simultaneous connections to the same
peripheral socket ID, for PML or any other service, so you're correct as we
discussed before that PML multiplexing will need to be handled by the host.
I suppose it's possible that a few of the high-end LaserJets with slots for
JetDirect cards support this, since they have to multiplex print and PML
between the parallel port and EIO/MIO slots anyway, but of course this
project isn't targeting those peripherals, so we needn't worry about that
case.  :-)

David