[interfacekit] Re: BSession(NG)

--- Marcus Overhagen <ml@xxxxxxxxxxxx> wrote:
> Pahtz <pahtz@xxxxxxxxxxxx> wrote:
> 
> > BTW, excuse me if you know already but read_port() is not like a socket or 
> > file descriptor.
> You
> > will lose data if the bufferSize is too small. You can't read_port() twice 
> > on the same
> message.
> Doesn't matter if you use fixed size buffers.

Personally, I don't mind if are saying:

limit on message size + fixed buffer of maximum size

It's just that no-one has stood up and said that :-)

> 
> > It's simple, it works. One malloc() call or none if the buffer is already 
> > large enough.
> You need to add one free() and one port_buffer_size() plus the realloc() 
> calls on the
> sender size, plus the overhead of dealing with variable buffer sizes.
> 
> I really don't understand why you propose this change. It has no benefit.

Again, if someone said it's unecessary to accomodate more 4096 bytes then I 
wouldn't propose it.

Why I am doing this is:

a) The BSession interface semantics are broken - it is used in hundreds of 
places and the
investment in this broken semantics will keep growing if we don't fix it
b) PortMessage and PortLink are two sides of the same coin
c) BSession and PortMessage/Link interfaces are wildly different
d) We have 2 protocols SESSION and PORTLINK when we could have one

i.e. we could have one class and one protocol instead of 3 classes and 2 
protocols

> 
> regards
> Marcus Overhagen



        
                
__________________________________
Do you Yahoo!?
Win a $20,000 Career Makeover at Yahoo! HotJobs  
http://hotjobs.sweepstakes.yahoo.com/careermakeover

Other related posts: