Sequential consistency implies a total order on writes, not just that processes see the writes in an order consistent with the order each individual writer used. This history is not valid according to sequential consistency. It _IS_ valid for the two processors to see different values, if one read is stale, but it wouldn't see a different order of writes.
Thank you for the feedback. When reading the literature I came to the impression that a single processor need not agree with others on which valid interleaving is observed. I could be wrong about this. Can you provide me any sources that helped you come to your conclusion? I will read thru and edit the post accordingly.
Should valid example 2 be y x z? I don’t understand the 2 y write processing so if I misinterpreted this, I would appreciate it if you could explain in more detail.
> Proc-4 Observes: W(y),W(x),W(z), R() => z
Sequential consistency implies a total order on writes, not just that processes see the writes in an order consistent with the order each individual writer used. This history is not valid according to sequential consistency. It _IS_ valid for the two processors to see different values, if one read is stale, but it wouldn't see a different order of writes.