Right, I don't think the actual memory usage is as big a concern as the interaction with the TCP stack. This could effectively force you to use an application-space buffer that may be far bigger than your TCP buffer, so you're just unnecessarily batching stuff up on top.
I agree this is almost *never* an issue in practice. I basically feel that way about this entire discussion, actually. -----Original Message----- From: Bryan Duxbury [mailto:[email protected]] Sent: Tuesday, May 04, 2010 11:49 AM To: [email protected] Subject: Re: heterogeneous collections On Mon, May 3, 2010 at 11:18 PM, Mark Slee <[email protected]> wrote: > also mandates the use of an internal buffer *at least* as large as the > largest complex object you are going to serialize > I think this is a bit of a red herring. In general, the size of the object you're trying to serialize, which must already be in memory, will dwarf the size of the object when it is serialized, except perhaps in very particular cases. You'd basically have to be operating at the edge of free memory already for it to be that much of an issue.
