servers-of-happiness is too conservative when K = 1

[davidsarah]

When K = 1, any single share is sufficient to reconstruct the file. However, the servers-of-happiness code will still require that there are H servers with unique share numbers. It should not care about the share numbers.

(We knew of examples with small K where the maximum matching definition of servers-of-happiness was conservative relative to the original definition, e.g. 778#comment:198. However, for K > 1 it is more space-efficient to use the maximum matching definition. For K = 1 it is not.)

[zooko]

I think we're out of time for 1.8.2. (Apologies if you're actually still intending to do this for 1.8.2.)

[davidsarah]

It makes sense to fix this at the same time as Kevan's changes to share placement behaviour in #1382, which will not be in 1.9.

[markberger]

For example, if we have a share distribution like this:

• Server A: [1]
• Server B: [1]
• Server C: [1]

with k = 1 and h = 3, the placement will be considered unhealthy even though we don't care about the share numbers because any single share is sufficient to reconstruct the file.