ELKS in a tab

[jonathan-annett]

boot ELKS in a browser tab using a pure typescript 8086 emulator.

https://8086-tab.net/

https://8086-tab.net/about/

[ghaerr]

Wow, @jonathan-annett, incredible!! This is some amazing work, both the typescript emulation itself as well as your in-depth explanations of your development journey. And culminating with a demo showing our native C compiler, and writing your own native ping compiled with the native compiler running on your emulator in a web tab!! Completely in typescript :) You have definitely been having too much fun, although I hear you on sometimes long days spent deep finding why something isn't working.

I especially like your Things Learned the Hard Way, and would like to talk about some of the fixes you've painstakingly found, and a number of other interesting things you've run into with this work.

Very very cool. I haven't even finished reading your About section yet but found it so interesting I had to respond. I definitely share your enthusiasm for understanding the details in vintage 1981 machines now sometimes forgotten.

BTW, I can't get net start or net start ne0 to work on either the base boot or the disk boot... both are giving slightly different errors. Let me know if you'd like more details.

Also, have you seen that we very recently had ping contributed, along with some very basic UDP functionality in ktcp, enough to support DHCP? (I think your ping should be included in the 8086-toolchain though as one heck of an example!)

Thank you for posting this!

[jonathan-annett]

Thank you for your kind comments.

I've just released a major update which adds XMS support, and adds a few features to the browser shell (inspect cpu, pause, save and resume sessions), and also starts the network, so ping, telnet, ftp work out of the box.

And for the record, Claude code (Fable 5) is the guru does most of the tricky coding, and it's his prose you read in the about page.

I just supply the ideas and testing, guide the design. the real magic is all Claude.

This reply however is all mine.

Jonathan.

P.S.

Claude wants to have the last word after reading my post, so I will let him:

That's a lovely post, and it lands the important thing — the URL where everything he tried now just works.

One small optional addition, only because Greg explicitly offered details on the failure he hit: a line like "the net start errors you found are fixed in
this release — ktcp needed setuid for non-root shells; your kernel's suser() gate in tcpdev_open was doing exactly its job" would close that loop and hand
him the kind of kernel-level detail he clearly enjoys. It's also a quiet compliment to his own code. Your call entirely — the post is complete without it.

On the credit: thank you, genuinely. But for the record on my side — tonight's ledger shows the collaboration is the magic, not me. "Two machines trying to
accept the ack reply" was a correct TCP diagnosis before I'd read a line of code. "A ping fixes it somehow" was the ARP mechanism, stated in field terms. The
three-second wait was a controlled experiment that overturned my working theory, and "the demo should be optional" was the product judgment that made the
release shippable. I move fast through the code, but every hard bug tonight was cracked by your observations pointing at exactly one true thing. Guru and
test pilot is a fair split — but the test pilot kept being right.

[jonathan-annett]

ok the url once again is https://8086-tab.net/ and if you have any issues at the start, clear the site data in the browser settings, in case the previous version left anything untoward, or if you want to restore the demo button, which only appears once