How to transfer data FAST?

[Johnny Billquist]

On Mon, 27 Jun 2005 Paul Koning <pkoning at equallogic.com> wrote:

> >>>>> "Johnny" == Johnny Billquist <bqt at Update.UU.SE> writes:
>
>  Johnny> True, and definitely a point for you.  I offered them access
>  Johnny> to systems to check it out, but no interest.
>
> I suppose that too makes sense -- if you don't know the OS, it's hard
> to get started on debugging communication to it.

No, I don't think that makes sense. If they knew enough VMS to test
against that, then RSX shouldn't be a problem. The syntax for fooling
around in DCL is the same.

>  Johnny> On a separate issue, the public domain LAT is equally
>  Johnny> problematic, and should not be used to connect to RSX
>  Johnny> hosts. You'll crash the RSX host eventually. That LAT
>  Johnny> implementation seem to produce some packets that RSX just
>  Johnny> don't deal with, which results in buffer pool getting lost on
>  Johnny> the RSX end. It's definitely not a good behaviour in RSX as
>  Johnny> well, but never the less, it appears as if the PD LAT breaks
>  Johnny> the protocol in some way, and LAT is even worse, since no
>  Johnny> documentation exists outside of DEC (or whereever), so it's
>  Johnny> all reverse engineering (once more of VMS).
>
> That's clearly a P1 RSX bug.  There's a simple rule: if your system
> crashes because of a packet it received, that is ALWAYS the fault of
> the receiving system.  Perhaps the sender should not have sent it, but
> "you shouldn't have sent that" is NEVER an acceptable excuse for
> crashing.

I totally agree. But that don't help much right now, since the LAT code in
RSX isn't available to us. So for now, the only options is not to use the
public domain LAT implementation against RSX.

	Johnny

Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email: bqt at update.uu.se           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol