-----------------------------------8<-----------------------------------
$ cat s.scm
(define p (open-tcp-client "news.tilde.club:119"))
(display (read-line p)) (newline)
$ ##
$ ## interpreted:
$ ##
$ gsi s.scm
200 news.tilde.club InterNetNews NNRP server INN 2.7.0 ready (no posting)
$ ##
$ ## compile:
$ ##
$ gsc -exe s
$ ls -l s
-rwxr-xr-x 1 yeti yeti 51404 Jan 23 06:59 s
$ strip s
$ ls -l s
-rwxr-xr-x 1 yeti yeti 10148 Jan 23 06:59 s
$ ##
$ ## run:
$ ##
$ ./s
200 news.tilde.club InterNetNews NNRP server INN 2.7.0 ready (no posting)
$ █
----------------------------------->8-----------------------------------

Maybe I really should play with a Scheme at least for a while.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

I neither found `(do ...)` nor `(while ...)`, but that's not a
showstopper for trying my standard bignum example:

-----------------------------------8<-----------------------------------
$ cat p.scm
(let loop ((p 1) (pp 1))
(set! p (+ p 1))
(if (= (gcd pp p) 1)
(begin
(set! pp (* pp p))
(print p " " pp "\n")))
(if (< p 42)
(loop p pp)))
$ gsi p.scm
2 2
3 6
5 30
7 210
11 2310
13 30030
17 510510
19 9699690
23 223092870
29 6469693230
31 200560490130
37 7420738134810
41 304250263527210
$ gsc -exe p
$ ./p
2 2
3 6
5 30
7 210
11 2310
13 30030
17 510510
19 9699690
23 223092870
29 6469693230
31 200560490130
37 7420738134810
41 304250263527210
$ █
----------------------------------->8-----------------------------------

Having interpreter and compiler is nice.

I tried Embeddable Common Lisp before and that was far more confusing to
use.
--
R || 0 ... Resistance is futile.
--- Synchronet 3.19a-Linux NewsLink 1.113

------------------------------------------------------------------------ (define l "")

(define p (open-tcp-client "news.tilde.club:119"))
(set! l (read-line p))

(display "ARTICLE <yay.scheme@87zfwwjzgd.xxx>\r\n" p) (force-output p)
(set! l (read-line p))

(let loop ()
(set! l (read-line p))
(if (not (string=? l ".\r")) ; read-line strips only the LF (of the CRLF)
(begin
(print l "\n") ; CR still in string :-/
(loop)
)
)
)

(display "QUIT\r\n" p) (force-output p)
(set! l (read-line p))

(display ".\r\n" p) (force-output p)
(set! l (read-line p))

(exit)
------------------------------------------------------------------------

Sometimes the simple tings take the most digging...
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Apropos too simple things:

------------------------------------------------------------------------
(print l "\n") ; CR still in string :-/ ------------------------------------------------------------------------

The `:-/` was `:-(` earlier and that did screw up the highlighting of
matching `()` in `mcedit`. So I could not make sense of my code
indentation versus the editor's `()`-counting.

Sometimes only partially smart is worse than dumb!

I really should start using `emacsclient` for fast edits instead of
`mcedit`.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

More digging through the manual starts to pay back!
Especially useful: `eol-encoding: 'cr-lf`
Now I have no longer to manually handle CRLF.

------------------------------------------------------------------------ (define println-to-port-and-flush ; /!\ needs nicer name!
(lambda (port string)
(println port: port string)
(force-output port)))

(define l "")

(define p
(open-tcp-client
(list address: "news.tilde.club" port-number: 119 eol-encoding: 'cr-lf))) (set! l (read-line p)) (println l)

(println-to-port-and-flush p "ARTICLE <yay.scheme@88zfwwjzgd.xxx>")
(set! l (read-line p)) (println l)

(let loop ()
(set! l (read-line p))
(if (not (string=? l "."))
(begin
(println l)
(loop)
)
)
)

(println-to-port-and-flush p "QUIT")
(set! l (read-line p))

(println-to-port-and-flush p ".")
(set! l (read-line p))

(exit)
------------------------------------------------------------------------
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

yeti <yeti@tilde.institute> writes:

------------------------------------------------------------------------ (define println-to-port-and-flush ; /!\ needs nicer name!
(lambda (port string)
(println port: port string)
(force-output port))) ------------------------------------------------------------------------

------------------------------------------------------------------------ (define (println-to-port-and-flush port string)
(println port: port string)
(force-output port)) ------------------------------------------------------------------------
--
( Are we there yet? )
\ ( Be patient! )
\ \
_______________________ _@o ____________ _@o _@o _______________________
--- Synchronet 3.19a-Linux NewsLink 1.113

After a lot of Lisp/Scheme blues (every dialect seems to handle sockets
in a different way) I remembered AWKA possibly having a FFI, so I could
access sockets somehow. Surprise, surprise: AWKA handles sockets
already like GAWK does.

------------------------------------------------------------------------
$ cat nntp.awka
BEGIN {
s = "/inet/tcp/0/news.tilde.club/119"
s |& getline
print $0
close(s)
}
$ awka -X -s -f nntp.awka
$ ./awka-app.out
200 news.tilde.club InterNetNews NNRP server INN 2.7.0 ready (no posting) ------------------------------------------------------------------------

Couldn't be easier, eh?

The size of the statically linked binary:

------------------------------------------------------------------------
$ ls -l awka-app.out
-rwxr-xr-x 1 yeti yeti 226576 Jan 31 13:33 awka-app.out
$ strip awka-app.out
$ ls -l awka-app.out
-rwxr-xr-x 1 yeti yeti 205264 Jan 31 13:33 awka-app.out
$ file awka-app.out
awka-app.out: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=7fa82163c9518cab94989c4a2b69a456401d6d1f, for GNU/Linux 3.2.0, stripped
------------------------------------------------------------------------

Dynamically linked:

------------------------------------------------------------------------
$ awka -X -f nntp.awka
$ ls -l awka-app.out
-rwxr-xr-x 1 yeti yeti 17696 Jan 31 13:40 awka-app.out
$ strip awka-app.out
$ ls -l awka-app.out
-rwxr-xr-x 1 yeti yeti 14536 Jan 31 13:40 awka-app.out ------------------------------------------------------------------------

As well known AWK fan looking for a language that shall fit well into
Nixens and interface nicely with C, AWKA really may be a candidate.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Well...

------------------------------------------------------------------------
BEGIN {
FS = "\r\n"
OFS = "\r\n"

s = "/inet/tcp/0/news.tilde.club/119"
s |& getline

print "article <87le85hatb.fsf@tilde.institute>" |& s
s |& getline

l=""
while( l!=".\r" ) { s |& getline l ; print l }

print "quit" |& s

print "." |& s

close(s)
}
------------------------------------------------------------------------

...works with GAWK, bu not with AWKA and I'm currently not in the mood
for a bug report.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Well...

------------------------------------------------------------------------
BEGIN {
FS = "\r\n"
OFS = "\r\n"

s = "/inet/tcp/0/news.tilde.club/119"
s |& getline

print "article <87le85hatb.fsf@tilde.institute>" |& s
s |& getline

l=""
while( l!=".\r" ) { s |& getline l ; print l }

print "quit" |& s

print "." |& s

close(s)
}
------------------------------------------------------------------------

...works with GAWK, but not with AWKA and I'm currently not in the mood
for a bug report.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Restart with v4.9.5...

`define-syntax` finally works:

------------------------------------------------------------------------
$ cat define-syntax.scm
(define-syntax while
(syntax-rules ()
((while test body ...)
(let wloop ()
(if test (begin body ... (wloop)))
))))

(define i 0)
(while (< i 10)
(print i " ")
(set! i (+ i 1))
)
(newline)
$ gsi define-syntax.scm
0 1 2 3 4 5 6 7 8 9 ------------------------------------------------------------------------
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Gambit Scheme 4.9.5 allows easier switching of the C backend:

------------------------------------------------------------------------
$ cat primorial.scm
(let loop ((p 1) (pp 1))
(set! p (+ p 1))
(if (= (gcd pp p) 1)
(begin
(set! pp (* pp p))
(print p " " pp "\n")))
(if (< p 42)
(loop p pp)))
$ gsc -cc gcc -exe -o primorial.gcc primorial.scm
$ ls -l primorial.gcc
-rwxr-xr-x 1 yeti yeti 19288 Feb 1 21:11 primorial.gcc
$ gsc -cc tcc -exe -o primorial.tcc primorial.scm
$ ls -l primorial.tcc
-rwxr-xr-x 1 yeti yeti 10960 Feb 1 21:12 primorial.tcc
$ ./primorial.tcc
2 2
3 6
5 30
7 210
11 2310
13 30030
17 510510
19 9699690
23 223092870
29 6469693230
31 200560490130
37 7420738134810
41 304250263527210 ------------------------------------------------------------------------

\o/
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

ChickenScheme's wiki is much more blingblingy than Gambit's manual.

<https://wiki.call-cc.org/man/5/Module%20(chicken%20tcp)>

But the answers for my Gambit questions were easier to find in Gambit's
less blingblingy manual.

That 1st shot at sockets in Chicken took definitely longer than the same snippet in Gambit:

------------------------------------------------------------------------ (define l "")

(import (chicken io) (chicken tcp))
(define-values (nntp-in nntp-out) (tcp-connect "news.tilde.club" 119))
(set! l (read-line nntp-in)) (print l)

(write-line "ARTICLE <yay.scheme@88zfwwjzgd.xxx>" nntp-out)
(set! l (read-line nntp-in)) (print l)

(let loop ()
(set! l (read-line nntp-in))
(if (not (string=? l "."))
(begin
(print l)
(loop))))

(write-line "quit" nntp-out)
(set! l (read-line nntp-in)) (print l)
(write-line "." nntp-out)
(set! l (read-line nntp-in)) (print l)
(exit)
------------------------------------------------------------------------

Not a beauty, just my 1st contact to Chicken.


Scheme may have some standard(s), but the interesting stuff like sockets
isn't covered in there. Gambit sockets yield readwrite ports, Chicken
two ports.
--
If you know one Lisp/Scheme, you know one Lisp/Scheme.
--- Synchronet 3.19a-Linux NewsLink 1.113

yeti <yeti@tilde.institute> writes:

ChickenScheme's wiki is much more blingblingy than Gambit's manual.

<https://wiki.call-cc.org/man/5/Module%20(chicken%20tcp)>

Omg, please translate ``blingblingy''. ``Shinning'', I say.

That 1st shot at sockets in Chicken took definitely longer than the same snippet in Gambit:

Compare it to C. :)

Scheme may have some standard(s), but the interesting stuff like sockets isn't covered in there. Gambit sockets yield readwrite ports, Chicken
two ports.

--
If you know one Lisp/Scheme, you know one Lisp/Scheme.

Very true. Now, what's even more interesting is why I don't exchange
Lisp for Scheme.
--- Synchronet 3.19a-Linux NewsLink 1.113

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

That 1st shot at sockets in Chicken took definitely longer than the same
snippet in Gambit:

Compare it to C. :)

I've a snippet that just spits out the NNTP server's welcome line, not a complete message (yet?). I'd like to have `defer` in C and easier/safer strings, but if I had to add reading a message via Message-ID, I'd not
fall into panic.

Unlike the dialect specific Scheme examples so far, the same C code
passes different C compilers. That's what should be underlined with a
flashy text marker.

-----------------------------------8<----------------------------------- #include <sys/types.h>
#include <sys/socket.h>
#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>

#define BUF_SIZE (4096)

int main()
{
int sock;

ssize_t nread;
char buf[BUF_SIZE];

struct addrinfo hints, *addrs;
memset(&hints, 0, sizeof(struct addrinfo));
hints.ai_family = AF_UNSPEC;
hints.ai_socktype = SOCK_STREAM;

if (0 == getaddrinfo("news.tilde.club", "119", &hints, &addrs))
{
sock = socket(addrs->ai_family, addrs->ai_socktype, addrs->ai_protocol);
if ( sock >= 0 )
{
if ( connect(sock, addrs->ai_addr, addrs->ai_addrlen) >= 0 )
{
/* Nothing to send, just read */
nread = read(sock, buf, BUF_SIZE);
if ( nread != -1)
printf("Received %zd bytes: %s\n", nread, buf);
else {
perror("read");
exit(EXIT_FAILURE);
}

}
close(sock);
}
freeaddrinfo(addrs);
}
}
----------------------------------->8-----------------------------------
( Works with GCC, PCC and TinyCC )

Very true. Now, what's even more interesting is why I don't exchange
Lisp for Scheme.

But from one Lisp to the next one, you'll probably not find things like
sockets handled in a standard way too.

Despite my frequent cursing about GCCisms breaking portable C code, for
stuff you write yourself, portability in C still seems doable. OTOH,
many huge projects like Emacs can't be compiled any more via compilers
without GCCisms and I'm very angry[0] about this and call this GNU/GCC
locked in syndrome ™Stallman's Revenge™.

____________

0: <https://www.youtube.com/watch?v=XlVoD17y5y4&t=4s>
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

yeti <yeti@tilde.institute> writes:

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

That 1st shot at sockets in Chicken took definitely longer than the same >>> snippet in Gambit:

Compare it to C. :)

I've a snippet that just spits out the NNTP server's welcome line, not a complete message (yet?). I'd like to have `defer` in C and easier/safer strings, but if I had to add reading a message via Message-ID, I'd not
fall into panic.

Unlike the dialect specific Scheme examples so far, the same C code
passes different C compilers. That's what should be underlined with a
flashy text marker.

#include <sys/types.h>
#include <sys/socket.h>
#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <string.h>

#define BUF_SIZE (4096)

int main()
{
int sock;

ssize_t nread;
char buf[BUF_SIZE];

struct addrinfo hints, *addrs;
memset(&hints, 0, sizeof(struct addrinfo));
hints.ai_family = AF_UNSPEC;

I would say you should say AF_INET here. POSIX.1 only specifies
AF_INET, AF_INET6, AF_UNIX. So your program is not POSIX.1 with
AF_UNSPEC.

hints.ai_socktype = SOCK_STREAM;

if (0 == getaddrinfo("news.tilde.club", "119", &hints, &addrs))
{
sock = socket(addrs->ai_family, addrs->ai_socktype, addrs->ai_protocol);
if ( sock >= 0 )
{
if ( connect(sock, addrs->ai_addr, addrs->ai_addrlen) >= 0 )
{
/* Nothing to send, just read */
nread = read(sock, buf, BUF_SIZE);
if ( nread != -1)
printf("Received %zd bytes: %s\n", nread, buf);
else {
perror("read");
exit(EXIT_FAILURE);
}

}
close(sock);
}
freeaddrinfo(addrs);
}
}

( Works with GCC, PCC and TinyCC )

Check all that bureaucracy! :-)

Very true. Now, what's even more interesting is why I don't exchange
Lisp for Scheme.

But from one Lisp to the next one, you'll probably not find things like sockets handled in a standard way too.

I had in mind things like error messages. SBCL's error messages are so
nice, for example. And Scheme virtual machines tend to be horrible.

Despite my frequent cursing about GCCisms breaking portable C code, for
stuff you write yourself, portability in C still seems doable. OTOH,
many huge projects like Emacs can't be compiled any more via compilers without GCCisms and I'm very angry[0] about this and call this GNU/GCC
locked in syndrome ™Stallman's Revenge™.

So far, nobody except friends have used my software. So to me writing
portable code have only helped with my education, but I think it's been
very much worth it. (For example, Writing portacle code is also a good exercise in abstraction.)

In practice, I view portability as an easy way to adapt your software to another system. Having it compile and run perfectly in ``all systems''
can only be achived with some specific types of programs if at all.
--- Synchronet 3.19a-Linux NewsLink 1.113

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

struct addrinfo hints, *addrs;
memset(&hints, 0, sizeof(struct addrinfo));
hints.ai_family = AF_UNSPEC;

I would say you should say AF_INET here. POSIX.1 only specifies
AF_INET, AF_INET6, AF_UNIX. So your program is not POSIX.1 with
AF_UNSPEC.

---<man 3 getaddrinfo>--------------------------------------------------
ai_family

This field specifies the desired address family for the
returned addresses. Valid values for this field include
AF_INET and AF_INET6. The value AF_UNSPEC indicates that
getaddrinfo() should return socket addresses for any
address family (either IPv4 or IPv6, for example) that can
be used with node and service. ------------------------------------------------------------------------

Sorry, I needed that IPv4/v6 tolerance.

Check all that bureaucracy! :-)

That allows plugging in more error checks in a way that fits the rest of
the application than just using one fat monolithic call.

For a quick throw away one liner I sure like it easier, e.g.:

---<manually \-wrapped for half a bit of readability>-------------------
$ gawk 'BEGIN { \
N="/inet/tcp/0/news.uni-stuttgart.de/119" ; N |& getline ; print \
; print "date" |& N ; N |& getline ; print \
; close(N) ; }'
200 news.uni-stuttgart.de InterNetNews NNRP server INN 2.5.2 ready (no posting) 111 20240323164058
$ # terminates, gives next prompt ------------------------------------------------------------------------

Near to unbeatable for some quick and dirty throw away code. For
standard AWK with some GAWKy extensions (e.g. that way of networking and coroutines), there's a compiler (AWKA), but unluckily original AWKA
doesn't even build currently and a different fork which builds hangs
after the NNTP welcome line and I sure wont start debugging foreign
code.

---<manually \-wrapped for half a bit of readability>-------------------
$ awka -x 'BEGIN { \
N="/inet/tcp/0/news.uni-stuttgart.de/119" ; N |& getline ; print \
; print "date" |& N ; N |& getline ; print \
; close(N) ; }'
200 news.uni-stuttgart.de InterNetNews NNRP server INN 2.5.2 ready (no posting) ------------------------------------------------------------------------

But from one Lisp to the next one, you'll probably not find things
like sockets handled in a standard way too.

I had in mind things like error messages. SBCL's error messages are
so nice, for example. And Scheme virtual machines tend to be
horrible.

Currently I'm looking at languages that compile to C. The mentioned
Schemes (and AWKA, BACON, Nim) do.

In practice, I view portability as an easy way to adapt your software
to another system. Having it compile and run perfectly in ``all
systems'' can only be achived with some specific types of programs if
at all.

I cannot check everything against GLIBC, MUSL, BSDCLIB, ... but I would
not use stuff I already know to be problematic.

But needing to handle sockets differently in every 2nd dialect of the
*same* language adds a level of not portable that's just scary.

I cursed about BASIC's and Pascal's diversity since TRS-80 and CP/M
days, so that's not really a new problem. And not even to mention all
those slightly incompatible assemblers for the same chip/CPU/ISA. Much
time got wasted over decades with such questions.

OTOH monopolies kill evolution.

I need more caffeïne!
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

Net access in this a bit queer (old meaning!) BASIC dialect (being
compiled to C) is very simple. When looking for a compiled alternative
to AWK, I played with it a bit, but when I reported a problem, they
slammed the door and that kind of killed all fun and trust. :-/

------------------------------------------------------------------------
OPEN "paganini.bofh.team:119" FOR NETWORK AS nntp
RECEIVE dat$ FROM nntp
REM no print, add error check later

SEND "ARTICLE <87wmvwubuz.fsf@tilde.institute>\r\n" TO nntp
RECEIVE dat$ FROM nntp
REM no print, add error check later

REPEAT
RECEIVE dat$ FROM nntp
PRINT dat$;
tail$ = RIGHT$(tail$ & dat$,4)
UNTIL tail$ == "\n.\r\n"

SEND "QUIT\r\n.\r\n" TO nntp
REM no print, add error check later

CLOSE NETWORK nntp ------------------------------------------------------------------------

Shit happens, moving on.
--
I do not bite, I just want to play.
--- Synchronet 3.19a-Linux NewsLink 1.113

yeti <yeti@tilde.institute> writes:

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

Patricia Ferreira <pferreira@example.com> writes:

yeti <yeti@tilde.institute> writes:

struct addrinfo hints, *addrs;
memset(&hints, 0, sizeof(struct addrinfo));
hints.ai_family = AF_UNSPEC;

I would say you should say AF_INET here. POSIX.1 only specifies
AF_INET, AF_INET6, AF_UNIX. So your program is not POSIX.1 with
AF_UNSPEC.

---<man 3 getaddrinfo>--------------------------------------------------
ai_family

This field specifies the desired address family for the
returned addresses. Valid values for this field include
AF_INET and AF_INET6. The value AF_UNSPEC indicates that
getaddrinfo() should return socket addresses for any
address family (either IPv4 or IPv6, for example) that can
be used with node and service. ------------------------------------------------------------------------

Sorry, I needed that IPv4/v6 tolerance.

I will ignore IPv6 until the whole world begins to use it. I might
leave IPv6 as heritage to my kids. :-)

Check all that bureaucracy! :-)

That allows plugging in more error checks in a way that fits the rest of
the application than just using one fat monolithic call.

Of course. It's C programming.

For a quick throw away one liner I sure like it easier, e.g.:

---<manually \-wrapped for half a bit of readability>-------------------
$ gawk 'BEGIN { \
N="/inet/tcp/0/news.uni-stuttgart.de/119" ; N |& getline ; print \
; print "date" |& N ; N |& getline ; print \
; close(N) ; }'
200 news.uni-stuttgart.de InterNetNews NNRP server INN 2.5.2 ready (no posting)
111 20240323164058
$ # terminates, gives next prompt ------------------------------------------------------------------------

Near to unbeatable for some quick and dirty throw away code.

The UNIX way.

OTOH monopolies kill evolution.

That's true. And diversity creates problems. Such is life.
--- Synchronet 3.19a-Linux NewsLink 1.113