Hello,

> Well, we've got a decent criteria list:
> 
> 1) mc0 performance > 10k/sec
Hmm, maybe find out why some people ( like me ) get halfway decent performance out of their mc's - Tim did some work on this and apparently found a fix.

> 2) serial port does not lock
Should be fixed with the new interrupt code ( if it works for you... ) - I transferred a couple of megabytes through tty01 at 38.400bps without a hitch.

> 3) devices behind bandit properly share interrupt
dito

> Separately,
> 
> 4) mesh works at 10 MB/sec
I get >8MB/s with bonnie++ - the latest -current version allows synchronous transfers ( you have to change the flags for mesh in your kernel config. The upper 8 bit control sync negotiation for each target - 1 means don't negotiate and is the default. Do NOT allow it for ZIPs though, it will hang for some reason )
mesh0 at obio0 offset 0x18000 (bootpath member)  irq 13: 50MHz, SCSI ID 7
scsibus1 at mesh0: 8 targets, 8 luns per target
...
probe(mesh0:0:0:0): max sync rate 10.00Mb/s
sd0 at scsibus1 target 0 lun 0: <SEAGATE, SX19171W, 9D23> disk fixed (bootpath m
ember) 
sd0: 8683 MB, 5268 cyl, 20 head, 168 sec, 512 bytes/sect x 17783112 sectors
probe(mesh0:0:3:0): max sync rate 10.00Mb/s
cd0 at scsibus1 target 3 lun 0: <PIONEER, CD-ROM DR-U03S, 1.01> cdrom removable
sd1 at scsibus1 target 5 lun 0: <IOMEGA, ZIP 100, J.03> disk removable
sd1: drive offline

> 5) multiprocessor machines don't lock
Indeed.

have fun
Michael