Ben, thanks for clarifying that. I had only glanced at the article
and noticed the author said he had to `cut a lot of corners' for the
project. It is an interesting project for remote monitoring with a
PIC. The normal approach is an embedded PC which drives up the cost.
I'm a novice when it comes to implementing TCP/IP.

   - Tom

At 12:05 AM 7/15/99 -0700, Ben Stragnell wrote:
>Hi Tom,
>
>The Circuit Cellar version doesn't implement TCP. It implements UDP/IP
>over PPP. In fact, it doesn't even implement a fully RFC-compliant
>version of that.
>
>Nonetheless, it's an impressive feat. This is why I *strongly* suspect
>that the device mentioned on slashdot doesn't do what the author claims.
>
>TCP is a *lot* (ie. order of magnitude) more complex than UDP, and,
>among other things, requires a copy of un-ACK'd data to be kept for
>resending (in the event of a lower-level IP packet going astray). Since
>the PIC sure as hell doesn't have space for this, is he storing this
>data on the EEPROM? Not a good idea if you have a limited write-cycle
>lifespan.
>
>I'm particularly impressed by his codesize breakdown - apparently
>HTTP/1.0 and I2C can be implemented in 3~99 instructions. Does this mean
>between 3 and 99 instructions? I'd like to meet the guy who can
>implement HTTP and I2C in 3 instructions. UDP+TCP in 70~99? If it's 99
>instructions, it ain't TCP (at least not in any meaningful sense).
>
>Cheers,
>Ben
>
>Tom Handley wrote:
>>
>>    Adam, I have'nt looked at the site but the current Circuit Cellar
magazine
>> (#108) has an article about a stripped-down TCP/IP stack implemented in a
>> PIC12C672 for remote sensing on the internet.
>>
>>    - Tom