Re: Open issues in the EXPUNGED draft

At 5:16 PM +0000 11/3/06, Alexey Melnikov wrote:

>  2). There was some debate recently about server's behavior if the 
> client provides a modsequence which is too old. Two solutions were 
> suggested:
>  a). The server returns all expunged messages from the specified UID 
> set (i.e. without paying attention to the provided modsequence)
>  b). The server returns "sorry, modseq is too old" and the client 
> needs to handle this.

Based on the earlier discussion, I'd say we go with option (a), since 
(from the earlier discussion)  since

At 12:06 PM +0200 9/26/06, Arnt Gulbrandsen wrote:
>  has two more advantages: The response is small on average. The 
> fallback needs no additional code in the client. (I love that. No 
> code to write, no extra case to test.)  And one disadvantage: The 
> worst-case response is big. The advantages are bigger, say I.

At 11:39 AM -0700 9/29/06, Randall Gellens wrote:
>  If the trade-off is really that the option that is simpler and 
> easier for clients to implement also generally has one less 
> round-trip and usually is just as compact, but in the worst case 
> can have a much larger response, I'm inclined to support it.  While 
> bandwidth saving is important, so is client simplicity.  Since we 
> expect that in most cases the response won't be larger, we don't 
> need to optimize for (hopefully rare) edge cases.

So, to summarize, let's optimize for the usual case and make