"J. Bruce Fields" <bfields@xxxxxxxxxxxx> writes: >> >> I'll resend it people are otherwise happy. > > I'm otherwise happy! Looks like a reasonable approach. > > (The one thing I wonder is whether it would be clearer to outright fail > writes that attempt to create already-expired cache entries. > > But I think that's an unimportant corner case really. And if there are > existing cases where that's happening then perhaps it's less disruptive > just to let them expire a second later rathern than to introduce a new > error.) I hadn't thought about. There is often a good case for failing a meaningless value rather than silently doing something different. Were I creating a new interface, I would probably do that. But revising an old interface that doesn't currently return an error... I agree with you that it is likely less disruptive to not introduce a new error. Thanks for asking the question. I'll post the proper patch separately. Thanks, NeilBrown
Attachment:
signature.asc
Description: PGP signature
