networking is relatively solid. one potential remaining issue is some limitation with full duplex async -- i'm not sure if shixin already has a fix for that. limitations/incomplete API depends on what you need; rather low-level TCP features are not exposed.
i don't know of any long-term connectivity issues with a _server_, but i only have stuff running with not too much traffic. no experience with long-term connectivity for a single connection -- maybe atronix has something to that effect running.
most _known_ random crashes in R3 are fixed. i think at the moment, of the _known_ crashes there's only two left: one a GC issue you'll not hit in practice with networking code, and the aforementioned PARSE problem with self-modifying rules that hits Lest (an issue that most likely exists in R2 as well, btw).
Maxim
thx Andreas.
on a related note: does R3 rebol-level api support half closing ports? selectively closing local or remote port for a single tcp link?
I've had problems getting some code to connect to common linux tcp tools in R2 because the servers expected us to close one port before they'd reply in the other. in R2, the close function always closes both ports.
Maxim
or rather, it always treats a closed port by forcing both sides of the connection to end, or treats it as such, even if the server is still responding on its half of the connection. many tools treat half closing the client side as the signal that you're done with your request.
(so the server can respond with its reply)
DocKimbel
About the datatype/word equality, I've considered that option for Red at the beginning, but thought that would bring more confusion than help, especially to newcomers. Datatype! is already difficult enough to grasp for beginners. Such equality introduces an annoying exception in word datatypes classification:
That said, I don't have a definitive opinion about it.
Geomol
Maybe it would be an idea to consider a datatype an any-word? Maybe the same for a none and a logic.
Gregg
There are times when that might make for less code, but I agree with Nenad that it may be more confusing.
Geomol
I think, many find SWITCH confusing when checking on types and forget to use TYPE?/WORD. Well, this with datatypes being equal to words with same spelling is in World, so it can be tried out in practise.
Maxim
SWITCH could consider datatypes (and the few syntactically ambiguous literal values) equivalent by default with an option to make it strict. I'd use the strict option maybe once a year... but I'd definitely like the default almost every day .
Maarten
Are there any docs on exactly what is in the Atronix builds, and what is the version/branch to use in terms of stability+features right now? I'm thinking of building and publishing a Docker image
szeng
Marrten, unfortunately, there aren't any doc on the features we added to atronix builds, but it includes FFI (similar (but more powerful) to R2's external function interface), encapsulation, CALL with I/O redirection, VIEW for Linux, and a lot of bug fixes
szeng
Adnreas, about full duplex async, I don't have a fix for it, and it's not my current focus, as it's a lot of work and might break compatibility with old scripts.
As for the long term connectivity, we don't see any issues, but we always issue a reconnection upon a broken connection, so we might haven't noticed it if there were some.
Maarten
szeng, thanks. Can you refererence PRs for me on the most imporant changes? I'll try and put something together - Atronix R3 deserves some more attention.
Maarten
Also, what are the GUI dependencies for view?
Pekr
Maarten - there should be no GUI dependencies, should be basically an AGG library ....
Maarten
There are some dependencies, but I've resolved them
There also seems to be an initiative making R3 C99 or so compatible, and packing TCC with it, which would make it self-compilable and maybe a bit closer to Red?