2017-04-27 13:26 GMT-04:00 Levente Uzonyi <leves at caesar.elte.hu>:

> Hi Ian,
>

Hi again,


> I see a few issues with the code you attached:
> - #receiveData will only read at most 2000 bytes and will get the content
> of one or maybe more TCP packets. But if you have larger packets enabled on
> your network, it'll only read a partial packet. I suggest you use
> SocketStream, because you don't want to process packets but a stream of
> data.
>

SocketStream seems like a good idea.


> - if you want to handle multiple requests through a single TCP connection,
> you should come up with a protocol to separate the different calls, so that
> you know where a request starts and ends. For example, you could just write
> 2 bytes, the length of the source string and then the source string itself.
> Or you could use Smalltalk's chunk format. Or whatever you come up with
> that fits your needs.
>

Yes, I was thinking about sending length of the source code first. That
would certainly allow SyntaxServer to know how many bytes to expect.

- locking is unnecessary for a single variable. As long as there's no real
> message send (non-failing primitives are not real sends) or backwards jump
> (loop) in your code, it'll be thread safe.


It was actually in a Pharo Sockets tutorial by Noury Bouraqadi and Luc
Fabresse.

Thanks for your feedback!

Ian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20170427/f638072f/attachment.html>