Friday, March 3, 2017

Archived a number of Alto cartridges, attended a working session at Xerox PARC, and made good progress on ethernet tool


Good work day today working on the Alto, archiving and verifying each cartridge on hand. The cartridges that I write from my tool are still not booting, however, even though my tool will read back the contents word for word.

Ken Shirriff's ethernet tool is now successfully reading every ethernet packet that the Alto receives, but flaky problems on his BeagleBone board make the transmissions bunch up all the manchester encoded transitions. Code that used to work is now failing the same way. He traced it to a timer which expires instantly rather than waiting for the programmed interval. He plans to get a second board and determine whether he has board problems

We rode over to PARC (the Palo Alto Research Center, a Xerox company) and had a great free form discussion among a number of PARC researchers who were involved in the Alto. On our side we also brought the CHM software curator, who hopes to create a center to study software on these systems.

Finally, we had a visit from John Shoch, a venture capitalist who spent 14 years at PARC, having been recruited by Alan Kay. John is now a trustee of the CHM and was president of a division of PARC before he moved to VC work.

Many great stories and lore, as we discussed ways to demonstrate Xerox Alto and possibly successor machines. Nothing firm but several interesting ideas were floated.

When we returned to Marc's lab, we brought nine more cartridges with us to archive. We had successfully archived the five we had on hand before this, but we have a strong discipline for new cartridges. We open them, clean them scrupulously and carefully examine the surface condition. This takes about half an hour per cartridge, time we didn't have tonight.

The last work we did was to collect some diagnostic information to figure out why the cartridges I write cannot be booted. We quickly found that some condition occurs that the Alto controller card does not expect and the read is aborted. It is while it is reading the preamble, a predetermined count of incoming words that should all be x0000, but the hardware is seeing a non-zero word.

Likely this is a timing issue, where my reading logic ignores the contents of words during the preamble so it happily detects the sync word and rest of the sector. I will be studying my notes and the logic in my disk tool to see if I can figure out what might be happening.

When we meet again, we will record the disk bit stream coming from the drive and compare both successfully booting packs and cartridges that I wrote which all fail to boot. If we can find the variations, I should be able to update the tool and write packs that work. 

No comments:

Post a Comment