Re: Using Memory BIO

2000-07-28 Thread Amit Chopra
> No, the record payload can be smaller than 16 KB and the record won't > be padded (except as far as necessary for the cipher). But each > record has a record header, and if you use smaller records, then a > larger percentage of network traffic will be record headers. A, my naivete shows. B

Re: Using Memory BIO

2000-07-28 Thread Bodo Moeller
On Fri, Jul 28, 2000 at 12:04:06PM +0530, Amit Chopra wrote: >> SSL_write will send the data passed to it in a single record unless >> unless the request is too large for one record. So for efficiency, >> avoid repeated calls to SSL_write with small (less than 16 KB) >> payloads if you can; if y

Re: Using Memory BIO

2000-07-27 Thread Amit Chopra
Hi, Thanks for the detailed explanation. Just a coupla more doubts :) Bodo Moeller wrote: > SSL_read cannot return anything until a complete record has been > received over the network. If the SSL_read request is for less bytes > than contained in that record, then the rest will be buffered.

Re: Using Memory BIO

2000-07-27 Thread Amit Chopra
Hi, Thanks a lot. I have seen your sources and ssltest.c and have picked up a few pointers from there. There are some general doubts in my mind though. Firstly , can WANT_READ/WRITE be issued during data tranfer or only during handshake ? A related question is does SSL_read/write do

Re: Using Memory BIO

2000-07-26 Thread Lutz Jaenicke
On Wed, Jul 26, 2000 at 10:03:09AM +0530, Amit Chopra wrote: > Hi, > I have a question about memory BIOs. > I want my application to SSL-read/write from/to memory > instead of sockets. Reading/writing to the sockets will be handled > by the application (I require such a facility because my