Toots for alcinnz@floss.social account

Written by alcinnz on 2025-01-22 at 00:51

Could it be that when we computer touchers were fighting our computers to get decent performance, that we got more say on the problems we tackled?

That deeper expertise made it harder to fire us?

=> More informations about this toot | View the thread

Written by alcinnz on 2025-01-21 at 23:44

I'm tempted to add an out-of-the-way control by which you can toggle daylight-savings or scroll through the timezones. That can be read in software via a specific memory address. Otherwise I'd stick with UTC!

Without downloading software updates from the net (I'm not interested networking for this machine), its not worth supporting political timezones or correcting for historical leap-seconds.

4/4 Fin for today!

=> More informations about this toot | View the thread

Written by alcinnz on 2025-01-21 at 23:39

Unless its given the string "*t" or "!*t" (in which case it returns that table), I'd interpret the 1st arg in a similar way to `string.format`. Substituting in the relevant fields.

As for timezone issues...

3/3.3!


=> /toot/113868980374910121 More informations about this toot
=> /thread/113868980374910121 View the thread

### Written by alcinnz on 2025-01-21 at 23:29

As for time...

If `os.time` is given a table, It'd feed that through a formula to a calculate a UNIX timestamp.

2/3?


=> /toot/113868943859413023 More informations about this toot
=> /thread/113868943859413023 View the thread

### Written by alcinnz on 2025-01-21 at 23:13
For escapism again... 2nd thread on boot-strapping Lua!
How'd we implement its `os` standard library without an underlying OS? Besides RISC-V's "Zicntr"?

To start I'd make several of these APIs, like `os.execute` & `os.getenv`, into noops! I see no reason to bother implementing a shell! As would `os.setlocale`, since (to avoid the need for more than 4GiBs of memory, both persistent & transient) I'd ship locales as cartridges to plugin next to the keyboard. That's how you set your locale!

1/3?


=> /toot/113868880314925779 More informations about this toot
=> /thread/113868880314925779 View the thread

### Written by alcinnz on 2025-01-21 at 22:30
That said: I'm not digging into details here to settle my stance, because I don't actually care about the OSCARs!


=> /toot/113868709025209447 More informations about this toot
=> /thread/113868709025209447 View the thread

### Written by alcinnz on 2025-01-21 at 22:28
I've run my own poll to judge this, but I trust Flipboard's more: https://flipboard.social/@CultureDesk/113862982823607408

It seems the fediverse is overwhelmingly anti-"AI". For most practical purposes so am I!

For this poll I think its utterly silly to take a hard "no AI in OSCARs contenders" stance. Though its fair to assess whether the UX of the tool more resembles commissioning or creating art.

Award-winning films were already making great use of machine learning, & other tech which in their time got labeled "AI"!


=> /toot/113868703875841391 More informations about this toot
=> /thread/113868703875841391 View the thread

### Written by alcinnz on 2025-01-21 at 18:55
I've pondered whether a decade or 2 ago we reached sufficiency for home computing.

This morning it strikes me that's about when the industry seems to have fully switched from innovating useful products to innovating new ways to profit. Could there be an explanation for this correlation?

Edit: "Fully" is an important word!


=> /toot/113867865767083330 More informations about this toot
=> /thread/113867865767083330 View the thread

### Written by alcinnz on 2025-01-21 at 18:38

We'd implement math.ult in Assembly to treat its integer argument as being unsigned. Assembly's multiply-add could be useful elsewhere.

As for math.random ... A handful of bitshifts & XORs should mostly do it, with a multiply to remove any obvious correspondance between consecutive calls!

3/3


=> /toot/113867796866470765 More informations about this toot
=> /thread/113867796866470765 View the thread

### Written by alcinnz on 2025-01-21 at 18:33

2/3?


=> /toot/113867776807576639 More informations about this toot
=> /thread/113867776807576639 View the thread

### Written by alcinnz on 2025-01-21 at 18:28
As for how I'd implement the rest of Lua's `math` standard library upon our Lua interpreter bootstrapped upon a risc32cfimv...

1/3?

=> More informations about this toot | View the thread

Written by alcinnz on 2025-01-21 at 18:13

Subclasses of this client & server are provided for ESMTP to add encryption & decent authentication.

A SenderMixin class alters both client/server implementations from the perspective of a single message, & with corresponding factory classes. There's classes for different SMTP authentication schemes, a function abstracting everything to send a message, & defines a new text encoding.

3/3 Fin for today!

=> More informations about this toot | View the thread

Written by alcinnz on 2025-01-21 at 18:06

... a widely used utility-method to send code-prefixed responses & wrappers thereupon, & received lines are dynamically dispatched to methods representing a state machine & in turn methods for each command. These commands gathers provided data, regex-parse & validates their input, & defers to surrounding to actually send these emails.

This class has a corresponding factory class, a reasonably similar implementation for the client-side with several template methods.

2/3!

=> More informations about this toot | View the thread

Written by alcinnz on 2025-01-21 at 17:55

Studying Ruffle's (Python networking framework) implementation of SMTP...

There's a utility for outputting RFC822-formatted dates, a counter/iterator for generating IDs used externally, "quoting" email addresses in triangular brackets, a class for regex-parsing email addresses, & a wrapper class representing users.

Core SMTP server class has a handful of properties, implements callback methods for managing connections & reporting errors to clients, customizable via template methods, ...

1/2?

=> More informations about this toot | View the thread