Hey ! This guy is great ! He did it before me ! And with the same thing I intended to use ! He must be using some black-magic to read people's mind :( Anyway, it's a good thing. But BTW, this isn't more than a TCP/IP stack with a bare telnet (to send a mail you telnet to the mail port, ...), but it is a great step :)
BUT, (1) Will he realease a LIBRARY so every one can use it ?
(2) Anyway I intended to use the same source (TinyTCP) for the Prosit networking layer... maybe I'll make the library myself.
(3) Does it work under LINUX ? I kinda hate Winblows more and more, as it crashes every day and then... And moreover Linux IS the Network-OS, wo WHY does he bother with DOS stuff ? Anyway, on long-term, it should be able to use PPP and connect to any good provider (who said AOL, I was talking about good ones, the ones that DOESN'T use proprietary stuff)... As with the 2nd question I may help (though I'm a bit busy), as I got a fully fonctionnal version of TiGCC (Yes, and it compiles the Prosit kernel, so... ;) )

1) I do not know. I was always NOSTUB programmer, but I agree that TCP/IP should be library. It is, question how to make it flexible. If someone else wants to do it, there is a source, of course, and I can translate my explaination how the thing works.

2) Good luck. Three main problems (that took few months to be aware) were memory allocation, odd memory addresses and 128 bytes buffer.

3) I have RedHat 5.2. But, I tried to establish SLIP with raised DTR and CTS, and it was not possible nor with dip, nor with slattach (which is not part of RedHat, BTW). PPP is really better option, but it requires more programming.

1)

The reason why I don't like NOSTUB (it is kinda euphemism), because:
- it does no allow the use of libraries, which implies:
- it doesn't allow libraries virtualisation, as they doesn't exist, which in turn implies:
- it doesn't allow these programs to be run under a multitasking environnment (Prosit for instance), something I find _very_ annoying :(
- they don't call UNIVERSAL (STANDARD) fonctions to perform actions (open a file, ...) so these non-existant functions cannot be hooked in to make them multitasking compatable.
- they all include everything from file access functions to grayscale interrupt, so
- they CANNOT be run under multitasking, as hardware ISN'T virtualisable without an MMU (Memory Managment Unit)
- even if the 'normal' (i.e. other) executable format isn't perfect, it allows to do some usefull things, which cannont be done with nostub progs (BTW, can nostub progs include a BSS segment ? I don't remember).
- it comes from TI, and depends on TI :(

To sum-up :
I don't like NOSTUB, because NOSTUB progs WILL NEVER be able to be multitasked :((( (It's anyway hard enough with Doors programs), so you can pray, but don't expect miracles :-(

2) Thanks !
Memory allocation is sometimes brain-damage as on every system with no MMU :(

odd memory addresses ? where ? every allocated bloc should start on a 16 bit-boundary anyway, don't they ?

buffers... well, Prosit already has a set of rotating-buffers primitives (duno if it can help here...) :)

3)
yes, PPP is maybe too complex for our little engines ;)
The guys at TI do have very strange ideas... :(

Arg. this forum skewed-up my indentation :(
Well, have a good sleep ! It's time for me Zz˛ :-)

1)
Basicly, you don't like simplicity because you want to have a load of things that you don't need on a calc?
yeah, i don't need games either, but multitasking is a little different

1) Well, virtualisation of system calls can be done also on NOSTUB programs, in easier or harder way.

AMS 2.05 allowed new method for function call,
via LINE 1111 call. In this case, all system functions are replaceable by simple change of this handler.

With older versions of AMS it is still possible to recognize ROM calls in programs we want to execute and patch them. It is enough to make copy of ROM jump table to our own table in RAM and replace instructions that read from it.

Finally, regular call of AMS windows dealing routines instead direct poking to video memory looks like suited for multitasking.

Ah, yes! Protection ports from HW 2, probabbly can be used for preemptive multitasking even with programs that poke to video memory. When task switcher gives job to background process, it will prohibit write to video memory. The handler will copy incrimined byte to somewhere else. When process became foreground, handler will be deactivated and video memory restored.

2) In function sed_IsPacket I added

if (((Longword)p_begin_packet)&1){
*p_begin_packet++;
}
/*Izbjeci neparne adrese*/
this did not exist in original source.

128 bytes buffer I mentioned is for received bytes from link port. Without writing own link port handler, it can not be resized.

why can't i use my Black link, i don't have a gray link. What would it take to make it so i could use a black link?

Write new server on PC side, or improve program that two calcs can communicate using TCP/IP.

In both cases there is a source, so it is not impossible.

However, staying on single server connection is not a point of my project. Black link cable can not be connected to standard modem. The final goal is to make TI89 a good alternative to PalmPilot, device that is useful for reading your mail or CNN news when you are away from home. Laptops are big and too usual for real hackers.

Kewl idea. Very impractical, but kewl. (Wireless connection - that's what I'm waiting for.) Sucks that the black link can't be used. Anyway, about the webpages not being able to show up...I was thinking about that. I don't have access to a server with Lynx installed right now (a text based browser which can be run from Telnet). Has anybody tried telnetting into a server with Lynx, running Lynx, and then connection to the net? Sounds like a possible solution to me.


Dark Ninja

BTW - I know the modem speed is probably 56k or somewhere close, but I know the calculator doesn't recieve the information that fast. Anybody have any idea as to how fast the information is retrieved to the calc?

With the Graph Link, the calc can receive at 9600 bits/sec (9K). If someone were to make a program with support for the parallel or serial homemade cables, that speed could increase to around 30Kbits/sec. TI-86 and below can only go up to 15 or 16Kbits/sec.

>TI-86 and below can only go up to 15 or 16Kbits/sec.

Um, no. The speed is utimitly limited by the processor. If any of you remember the original expander then you know that Mel Tsai got the 85 to transmit at almost 50kbs!

Rogue

Romain Lievin's FastAVRLink goes as fast as 57600 bps...
and it should work here as it emulates the Graphlink cable :)

(and hey ! it hould work with Linux as-is as it can be powered by an external batery ;) )

I was quoting some figures I found a while ago on ticalc.org - and the 15 or 16 kbits/sec is what I clocked with FastLink (which I wrote) using the serial and parallel cables. The actual maximum may be somewhat higher, but definitely not 50k unless you overclock the calculator or bypass the TI-OS link routines and make your own.

If you look at the benchmarks Romain gives on his webpage, the transfers don't reach 56k. The best transfer for the TI-92 worked out to 35.6 kbits/sec, and the screenshot transfer for the TI-82 (I'm not considering the backup - there's something wrong with those figures - a TI-82 backup does not have 51 sections and the TI-GL does not work at 12k) worked out to 23.6 kbits/sec.

So yes, the speed is limited by the calculator's processor (6MHz for TI-82 to 86, 10MHz for 89 to 92+). But no, the transfers do not reach 56k without modification of the calculator hardware or software.

Waiting for wireless? Why not use the radiolinc 1? Get a grey cable and attach the calc end to the radiolinc. Wireless internet (I don't know if this works, the radiolinc is too expensive for me to buy, but I don't see why not!)

Me and my friend ordered a set of the Radio Links. We got them about 2 weeks ago and still haven't gotten them to work yet. The instructions are horrible and for some strange reason certain things on them are blacked out (like their phone number...). My friend is thinking that he has it figured out and is still working on the soldiering. We read from Radio Links' website that it was 3 wires to attact to the calculator. That is so wrong!
If you were going to be some, I would wait. Unless you think your smarter then us and can figured it out. I don't care how you waste your money.

heh, i alone bought two radiolinc v2.....cost me out the ass, and yes the instructions are terrible, but i have figured them out and if you need help, mail me or IM me

-garrett

I am BeefGG's friend that has figured it out. I was just wondering how you wired your power source to yours. I used four N size batteries to create a 6 volt power pack. I think that I will post pictures of my finished product on the web so others can see it (if it works). I will email you them if you want.

P.S. - There is no freakin way that these can be internal.

As far as I know, newer versions of NOS support also HTTP server. I have not yet downloaded it (I always forget address).

However, I hope that someone will solve problem with Linux SLIP driver. Even more, if someone reports to me that he successed to connect to provider using combination TI+gray graphlink+gender changer+null modem+modem+telephone line I will be very happy. Then remains the most interesting step, building TI89-GSM interface.

This is awesome. Does anyone have plans for a ethernet adapter cable for it yet?

Adapter cable ? when you can have a full 10baseT ethernet card ?
see http://www.embeddedethernet.com/ (or click the URL above)

Original TinyTCP contains also driver for one Ethernet card and ARP protocol. I removed this part from source, because TI89 does not have Ethernet card yet.

Not yet ! ;)

Yeah, it does have the ethernet driver :)
I have studied the original code for a bit now, but I still don't have the time I'd like to spend on Prosit and TI-programming :(