View Single Post
Old 19 April 2017, 09:24   #57
Olaf Barthel
Registered User
Join Date: Aug 2010
Location: Lehrte, Germany
Posts: 141
Originally Posted by gulliver View Post
Amiga TCP/IP stacks are not great by any means. They arent worth saving IMHO.

They dont support any modern feature, and within time they will be rendered useless (no IPv6 support).

There are lots of very good open source lightweight TCP/IP stacks which could be ported and perform much better with more chance of being future proof.
Sorry for the technical angle, but getting a TCP/IP stack working on the Amiga, keeping it working with the existing TCP/IP client software, is no mean feat.

The TCP/IP stacks which work with the Amiga client software are typically ports of the BSD kernel code, and this means that almost the entire kernel API existing around 1991 has to be supported, or your client software won't work (or will fail in interesting ways).

The typical lightweight TCP/IP stack doesn't have to support that much of an API, and replicating it is a real challenge. For example, some stack support poll(), but not select(), or select(), but not poll(). Rewriting a poll() or select() emulation is not a trivial exercise.

Because the Amiga has had Unix kernel TCP/IP software right from the start, back in 1987, porting it required cracking a number of challenging problems. This begins with making the TCP/IP stack's internal memory management work.

Why I am writing about all of this? You can't consider porting any modern TCP/IP stack to the Amiga (don't suggest writing one from scratch, please) without knowing about how one might be integrated on the Amiga platform. So, I believe it's important to save that code.

Incidentally, AmiTCP 3.0 is still on Aminet, and it's covered by the GPL.
Olaf Barthel is offline  
Page generated in 0.06099 seconds with 9 queries