topic Re: VMS731_LAN-V1200 in Operating System - OpenVMS

VMS731_LAN-V1200

Ian Miller. — Thu, 21 Oct 2004 03:32:31 GMT

I was just reading the release notes (just for fun :-) and I noticed some interesting points.

It is possible for ethernet frames to be reported as successfully sent when the driver has not sent them yet. The failure to send will only be reported in the statistics !

It is possible for a frame to be reported as not send due to timeout and infact it gets successfully sent !

As it points out, the transmission of data in ethernet frames is a datagram sort of service and higher level protocols have to cope.

One bit good news is that this patch kit has the duplex mismatch detection code in (mentioned at the TUD) - a console message is output by the driver if it thinks there may be duplex mismatch (e.g NIC on full duplex, switch on half uplex). This should help detect this problem which has bothered some people here.

Re: VMS731_LAN-V1200

Uwe Zessin — Thu, 21 Oct 2004 04:13:31 GMT

Oh, cool. The first one sounds a bit like a writeback cache on the network ;-)

Re: VMS731_LAN-V1200

Ian Miller. — Thu, 21 Oct 2004 04:32:57 GMT

and writeback cache without battery backup is an abomination unto Nuggen !
(see Terry Pratchett Monstrous Regiment)

Re: VMS731_LAN-V1200

Uwe Zessin — Thu, 21 Oct 2004 05:55:34 GMT

... but it saves money which is more important ;-)

Re: VMS731_LAN-V1200

Jan van den Ende — Thu, 21 Oct 2004 08:01:29 GMT

Ian,

see Terry Pratchett

Shouldn't information going to The Unseen University be shielded for commoners?

Looks like they achieved it :-)

Cheers.

Have one on me.

Jan

Re: VMS731_LAN-V1200

Martin P.J. Zinser — Thu, 21 Oct 2004 17:01:27 GMT

Hi,

well the first paragraph makes one wonder how the DEATH of network pakets looks like and what he uses as a scythe ;-)

Sorry, could not resist, it is too long I read my last Pratchett book.

Cheers,

Martin

Re: VMS731_LAN-V1200

Ian Miller. — Sat, 23 Oct 2004 13:55:17 GMT

All this humour does lighten up this dark and wet night (as does the beer I'm drinking - Belgian abbey ale called Maredsous - a triple at 10% - light in colour and very good).

However the point of my posting was to see if people thought that the behaviour of the LAN driver with this patch was surprising. It can report success when the transmit has failed and failure when the transmit succeeds! This seems very strange to me although I do accept the point about it being a datagram level service.

Re: VMS731_LAN-V1200

Bojan Nemec — Sat, 23 Oct 2004 15:06:28 GMT

Ian,

Yes, it seems strange. But if you think abbout collisions, packets droped by network devices (bridges,switches,routers...) and packets which are not received in sequence, this must be handled by a higher protocol. Which is the difference between a not sent packet and a droped packet?
Maybe the sent packet reported as not sent is slightly more difficult to handle. I dont know how higher protocols handle this. TCP puts a sequence number in the packet header and probably drops duplicate packets.

The duplex mismatch detection code is realy a good new. This is realy a problem which is seen to much times. The worth thing abbout this problem is that the network works, but like an old modem connection.

Bojan

Re: VMS731_LAN-V1200

Wim Van den Wyngaert — Sun, 24 Oct 2004 10:58:25 GMT

Ian,

Good choice, Maredsous. Also try Chimay blue and Rochefort 10.

1) It is possible for ethernet frames to be reported as successfully sent when the driver has not sent them yet.
-> Yes. You get an OK as fast as it is put into a buffer. It is the tcp or decnet software that will receive the NOK and increase the counters/do retransmit.

2) It is possible for a frame to be reported as not send due to timeout and infact it gets successfully sent !
-> Yes. If the network fails when your frame was sent and a network failure occurs before the ack/answer is received.

Wim