2022-12-08 03:55:55 <veltas      > Was -4*C this morning driving to work
2022-12-08 03:56:50 <boru        > Were you in warp drive or something?
2022-12-08 03:57:05 <boru        > -4 * C sounds like some serious hyperspace shenanigans.
2022-12-08 04:04:34 <olle        > Hehe
2022-12-08 04:05:03 <olle        > He took the shortest path straight through a black hole :D
2022-12-08 04:14:19 <veltas      > I think at this point I would have saved time looking up the degrees symbol
2022-12-08 04:14:53 <boru        > And denied us the opportunity to poke fun? For shame.
2022-12-08 04:15:19 <veltas      > Actually my message is related to today's XKCD
2022-12-08 04:15:37 <boru        > Ah, I see. I don't read it.
2022-12-08 04:16:07 <olle        > He still makes those?
2022-12-08 04:16:23 <veltas      > Not on purpose though
2022-12-08 04:16:52 <boru        > Cringeworthiness springs eternal.
2022-12-08 04:17:02 <veltas      > Every 5th one is sort of funny, it's always been the way. Ones that aren't funny are at least interesting usually
2022-12-08 04:17:22 <veltas      > It's probably the least cringey webcomic though
2022-12-08 04:17:25 <veltas      > The bar is not high
2022-12-08 04:18:10 <olle        > Hehe
2022-12-08 04:18:14 <olle        > Oatmeal t ho
2022-12-08 04:18:15 <boru        > There used to be one or two clever ones, mostly dark humour, but they're also prolific with it these days, as well.
2022-12-08 04:18:15 <olle        > tho
2022-12-08 04:18:33 <boru        > Oatmeal is hit or miss. Same with SMBC.
2022-12-08 04:18:57 <boru        > Though I haven't read either in a few years now.
2022-12-08 04:19:55 <boru        > Wondermark can still be funny occasionally.
2022-12-08 07:42:45 <olle        > oh noe
2022-12-08 07:42:53 <olle        > 64 nicks here, 300+ nicks in #lisp D:
2022-12-08 07:43:16 <olle        > Lemme troll them and see
2022-12-08 08:06:40 <veltas      > We don't still have a nick limit I hope
2022-12-08 08:14:00 <crc         > no nick limit here
2022-12-08 08:54:25 <veltas      > Thanks
2022-12-08 08:54:36 <veltas      > Sorry I forget how to check, I know can be done with ChanServ
2022-12-08 09:20:43 <KipIngram   > So, the other day I mentioned the possibility of keeping things like strings and arrays in a "ring buffer" that wrapped around on itself, and didn't really have explicit notions of garbage collection and other "management" associated with it.  Correct operation would be achieved by simply making it "big enough."  Are there any extant languages that operate in anything like this way?  I'm not aware of any, and
2022-12-08 09:20:45 <KipIngram   > given that it doesn't have "provable correctness" (you could always define a problem large enough to break it), I figure there may not be.  On the other hand, sometimes I think everything has been tried at some point or another.
2022-12-08 09:21:49 <thrig       > 640 processor registers ought to be enough for anyone
2022-12-08 09:22:45 <KipIngram   > Is that a joke, or is there actually a processor with that many?  Yeah - one would think that would cover most needs.
2022-12-08 09:23:22 <thrig       > registers do not really have GC and provided there are enough of them ...
2022-12-08 09:25:47 <KipIngram   > Well, it's slightly different, though.  A register will always have the last value you put into it.  But this setup I'm picturing will place large structures in that buffer and hand you back an address to it on the regular stack.  For a while, that value will be good, but once the buffer wraps and re-uses that RAM, you might still have that address on the stack but it's totally bogus now.
2022-12-08 09:26:06 <KipIngram   > The idea would be to be done with whatever you were doing before your pointers became stale.
2022-12-08 09:26:32 <thrig       > it's fine, until it isn't (like many things in computing)
2022-12-08 09:27:46 <KipIngram   > Sure.  I think it could be made to work ok for a lot of applications.  I just wondered if there are any systems that take that approach.
2022-12-08 09:32:24 <thrig       > I've looked at ring buffers for game energy systems and ring buffers while fast are pretty inflexible so if you need something more complicated uh rewrite
2022-12-08 09:37:13 <eris[m]     > <KipIngram> "Is that a joke, or is there..." <- one exists with 1024
2022-12-08 09:37:22 <KipIngram   > Yeah - the payoff of it would be that really almost no management overhead would be required.  I guess it's an extension of the PAD idea - PAD is using unallocated RAM.
2022-12-08 09:37:32 <KipIngram   > Wow, that is quite a lot.
2022-12-08 09:37:54 <KipIngram   > Are they individually referenced by name or index, or do they have some sort of "automation" layered on them?
2022-12-08 09:38:39 <KipIngram   > Or, I guess, do they present a standard register interface, or do you treat it like a small cache?
2022-12-08 09:39:41 <eris[m]     > both
2022-12-08 09:39:47 <eris[m]     > cant remember the name now
2022-12-08 09:44:18 <KipIngram   > Anyway, what's motivating me here is just to be able to deal with data that won't "fit" on the stack, while keeping the "operational flavor" of the system as close to normal Forth as possible.
2022-12-08 09:45:43 <KipIngram   > But I also want the operation of it to work with that ropes memory management I've talked about as well.  Steps on a path.
2022-12-08 09:47:03 <KipIngram   > Adding the ropes would mean explicitly managing the allocation and deletion of items, which would allow items to stick around long term if I needed them to.  But otherwise it would  "work the same."
2022-12-08 09:48:43 <KipIngram   > I think having 64 bit processors makes this more attractive - that's enough to let you use some bits of an address item to carry other information.  On narrower processors it would chew up an awful lot of your address space.  64 has room to spare.
2022-12-08 09:49:47 <KipIngram   > Might urn into trouble, depending on how hte OS you're running under manages the address space.  I don't know - maybe Linux uses some of those high bits for something.
2022-12-08 09:50:23 <KipIngram   > Just because I only have 16 GB of RAM doesn't mean that my addresses are all zero above bit 33.
2022-12-08 09:51:07 <veltas      > 6502 sort-of has 128 pseudo-registers
2022-12-08 09:54:59 <KipIngram   > I guess what defines the difference between a "register system" and a "memory system" for me is that register specifications are hard-coded into machine instructions, and that there's no notion of a "pointer" to a register.
2022-12-08 09:55:10 <KipIngram   > Not just fast vs. slow.
2022-12-08 09:55:42 <KipIngram   > I.e., the registers have no "adjacency" characteristics.
2022-12-08 09:58:03 <KipIngram   > If someone wanted to, they could have their registers just be the first few cells of RAM, and provide a mechanism for "short addresses" that instructions could use.  So everything would look like a RAM reference, but you'd still get more compact instructions and faster hardware speed when you referenced those instead of addresses that went to physical RAM.
2022-12-08 09:59:11 <KipIngram   > That would actually make for some interesting addressing modes - you always want a way to use the value in a register as an address, but once you had that then you could use the value in any RAM cell as an address.  The instruction would just be bigger if you used a "real RAM" address.
2022-12-08 10:01:32 <veltas      > Well you can hard-code an address into an instruction
2022-12-08 10:01:48 <veltas      > And hard-code a byte address into 6502 instructions
2022-12-08 10:02:07 <KipIngram   > Yeah; it just seems like the whole thing would be a little more "unified" with the setup I just described.
2022-12-08 10:02:09 <veltas      > It's all a bit arbitrary IMO
2022-12-08 10:02:23 <KipIngram   > It is, for sure.  Hardware can be designed to do damn near anything.
2022-12-08 10:02:37 <veltas      > Most arch's have a "short address" mechanism like you've described
2022-12-08 10:02:47 <veltas      > And 6502 had such a mechanism too for the zero page
2022-12-08 10:02:58 <KipIngram   > Yes; I guess I was thinking of a "really short" one.  :-)
2022-12-08 10:03:05 <KipIngram   > Like five bits or soemthing.
2022-12-08 10:03:08 <olle        > KipIngram: The normal way is probably to use a big enough pool instead of a ring buffer
2022-12-08 10:03:12 <thrig       > I hate it when addresses short out
2022-12-08 10:03:44 <olle        > At least, I can't see any obvious benefit to a ring compared to pool/region
2022-12-08 10:03:54 <KipIngram   > Well, with a ring you never "run out."  Basically it would be treated like a pool.
2022-12-08 10:04:24 <olle        > In both cases you "must" know you upper bound/worst case
2022-12-08 10:04:34 <olle        > But a pool can abort
2022-12-08 10:04:39 <thrig       > probably depends on how bad the overwrites or blocking until something frees up problems are
2022-12-08 10:04:40 <olle        > Abort > silent error
2022-12-08 10:04:56 <veltas      > Silent error is more forthy
2022-12-08 10:05:02 <KipIngram   > Sure, and with a ring you would have to watch out for whether the item you're processing now will fit in "the last part"; should you wrap  now or wait?
2022-12-08 10:05:19 <KipIngram   > I wouldn't likely try to allow a data item to internally wrap.
2022-12-08 10:05:41 <KipIngram   > Because then I'd have to consider that possibility every time I dealt with any item.
2022-12-08 10:06:06 <olle        > veltas: X)
2022-12-08 10:06:41 <olle        > I guess a ring could be more performant if upper bound is indeed known
2022-12-08 10:06:49 <KipIngram   > So I'd probably want to parse items fully, so I knew how big they were, before putting them in the ring.  Or else be prepared to do some kind of "Oops, I'm out of space; copy the first part of this one back to the bottom" sort of thing.
2022-12-08 10:06:51 <olle        > And it's not too different from average case
2022-12-08 10:07:09 <olle        > Then again, to free a pool is O(1), no?
2022-12-08 10:07:40 <thrig       > probably really good in a simple system where you can prove, block, or don't care about overwrites
2022-12-08 10:07:41 <KipIngram   > Yah, a simple fixed size memory manager isn't that expensive.
2022-12-08 10:08:03 <thrig       > could be the very wrong abstraction when those do not hold
2022-12-08 10:08:17 <veltas      > Unfortunately there's never a guarantee you can use specific address ranges, even on Linux. Can still do what you describe allocating pages though, just need to map the page number somewhere
2022-12-08 10:08:49 <veltas      > I agree with olle in principle
2022-12-08 10:08:59 <KipIngram   > Yeah, the issue with address bits comes in not from having a ring, but from attempting to "tag" stack items with type-like information in the high bits.
2022-12-08 10:09:19 <KipIngram   > separate concept
2022-12-08 10:10:23 <KipIngram   > The way I'd like for that to work would be to be able to take such an item, mask off the top eight bits or something, and have the address of the object.
2022-12-08 10:11:53 <KipIngram   > And whatever the "normal" state of those high bits is (I've usually assumed it's zero, but it's really whatever your system gives you), that would be the "type" for integer.  So that normal  Forth operation would automatically give you a correct type field up there for regular stack items.
2022-12-08 10:12:57 <KipIngram   > Well, that requires those bits be zero, I guess.  You'd use tag 0 and tag FF both to specify "integer," so that both positive and negative integers "looked right" to the type system.
2022-12-08 10:13:54 <KipIngram   > Anyway, that's a digression from the earlier topic.
2022-12-08 10:16:58 <olle        > You wanna box your stuff?
2022-12-08 10:17:13 <KipIngram   > ?
2022-12-08 10:17:22 <olle        > Box types for gc
2022-12-08 10:17:25 <olle        > boxed values
2022-12-08 10:18:33 <KipIngram   > Oh.  Well, I don't know what that entails really; I've just tried at times to think of a way to extend Forth to be able to do things similar to Octave.  Manipulate vectors and arrays, primarily, in some graceful fashion.
2022-12-08 10:18:56 <KipIngram   > These days that entails having stack items carry the address and type of the objects they reference, and the items themselves would live in some sort of heap.
2022-12-08 10:19:18 <KipIngram   > The ropes stuff and the ring buffer are just alternate ways of providing that heap, with different effort/performance tradeoff.
2022-12-08 10:20:02 <KipIngram   > In past years I've also thought about having a separate, explicit "type stack," but at the moment I'm more focused on trying to keep the information in the normal stack along with the values.
2022-12-08 10:21:27 <olle        > OCaml marks ONE bit for boxed vs unboxed
2022-12-08 10:21:40 <olle        > So ints are 31 bits I think
2022-12-08 10:25:23 <KipIngram   > Interesting.  I gotta shove off for the office - back later on.
2022-12-08 10:25:46 <olle        > \o
2022-12-08 10:40:43 <olle        > Is 8th needed?
2022-12-08 10:40:55 <olle        > Not sure why implementor didn't work on "normal" Forth instead
2022-12-08 11:34:04 <crc         > Ron (the developer of 8th) used to develop a more traditional Forth called Reva
2022-12-08 11:34:22 <crc         > https://ronware.org/dev/p.sh/reva/home
2022-12-08 11:38:54 <gordonjcp   > I thought 8th was "like forth but not forth"
2022-12-08 12:07:14 <olle        > crc: that still doesn't explain why not use an existing system
2022-12-08 13:24:34 <crc         > 8th has things (like actual data types) that break with traditional forth
2022-12-08 13:29:01 <crc         > two forum posts re the background & differences are at https://8th-dev.com/forum/index.php/topic,714.0.html and https://8th-dev.com/forum/index.php/topic,94.0.html
2022-12-08 14:45:03 <lagash      > What's the copyright(s) for Vierte Dimension anyways? Is it libre? Or they just made their archives open to all?
2022-12-08 15:15:01 <olle        > crc: You can't implement that inside classic Forth?
2022-12-08 15:21:58 <olle        > crc: Meh, those pages look like ad for Java ><
2022-12-08 15:22:30 <veltas      > You could implement it inside a forth but that's just yet another layer
2022-12-08 15:22:35 <veltas      > Not good for performance etc
2022-12-08 15:24:16 <olle        > Better than splitting the eco-system once again... :|
2022-12-08 15:24:48 <veltas      > Well Forth's incredibly split, you either do standard forth or you do whatever you want
2022-12-08 15:25:02 <olle        > That's sad
2022-12-08 15:25:03 <veltas      > And people do
2022-12-08 15:25:18 <olle        > No strong leadership?
2022-12-08 15:25:21 <veltas      > Well even Chuck Moore doesn't do standard forths
2022-12-08 15:25:42 <veltas      > I suppose part of the problem is he went off to focus on hardware for a while
2022-12-08 15:26:15 <olle        > :d
2022-12-08 15:30:06 <crc         > I use a non-standard dialect of Forth; I wouldn't be happy using an ANS Forth
2022-12-08 15:30:40 <veltas      > Yeah part of the problem with the standard is that it isn't good, it has tons of history in it and the new stuff doesn't really fit with the old that well
2022-12-08 15:30:57 <veltas      > I don't think the standards have ever done a good job unifying what people were doing, even back in the old days
2022-12-08 15:31:13 <veltas      > Which is why the early standards contain a lot of stuff that never made it into ANS
2022-12-08 15:32:06 <veltas      > But there is an aspect to forth of creating your own language, so it's natural that many dialects exist and is sort of intentional
2022-12-08 15:32:13 <veltas      > Forth is more of an idea than a language
2022-12-08 15:34:02 <olle        > Sucks
2022-12-08 15:34:16 <olle        > You pick a language because of the eco system
2022-12-08 15:34:36 <veltas      > Well you either work with standard Forth or you don't
2022-12-08 15:34:51 <veltas      > Just because people do non-standard forths doesn't mean you can't stick within the standard ecosystem
2022-12-08 15:35:04 <olle        > Kinda besides the point
2022-12-08 15:35:56 <crc         > I use my own ecosystem
2022-12-08 15:36:04 <olle        > Everyone does xD
2022-12-08 15:36:05 <veltas      > For standard forths there's a website for forth packages, not sure what that's like
2022-12-08 15:36:07 <olle        > In Forht-land
2022-12-08 15:36:15 <veltas      > There's the gforth community, there are the corporate forths
2022-12-08 15:36:44 <veltas      > That's all standard (or rather supports standard forth, they all extend it as well)
2022-12-08 15:36:55 <veltas      > My forth for the ZX Spectrum is ANS standard(ish)
2022-12-08 15:37:03 <veltas      > There's loads of standard stuff
2022-12-08 15:37:13 <veltas      > So let people do non-standard stuff, it's up to them
2022-12-08 15:38:14 <veltas      > If you want to try and work to build up the largest forth group then yes, use the standard
2022-12-08 15:38:48 <veltas      > Not everyone has that goal, it's my goal, I don't mind the old-fasioned forth style so I stick to that with what the standard supports
2022-12-08 15:45:01 <olle        > Ya sure
2022-12-08 15:45:09 <veltas      > olle: It's not too hard to use C stuff in gforth and some other standard forths, so your 'ecosystem' is the largest one there is ... the C ecosystem
2022-12-08 15:45:18 <olle        > Well
2022-12-08 15:45:18 <veltas      > If you want it
2022-12-08 15:49:08 <veltas      > olle: And then another point is that if you understand "the forth way" you know you don't *need* an ecosystem for a lot of worthy tasks anyway
2022-12-08 15:49:28 <veltas      > Just write your own code, keep it simple, minimise the requirements
2022-12-08 15:50:18 <olle        > I've heard that :) But it's a death sentence for a language's growth potential
2022-12-08 15:50:30 <olle        > Doomed to be esoteric
2022-12-08 15:50:37 <veltas      > Forth is actually growing though, recently
2022-12-08 15:50:51 <olle        > All languages are growing, because devs are getting more
2022-12-08 15:51:05 <veltas      > I don't doubt it won't become a top 10 language
2022-12-08 15:51:14 <veltas      > And neither will its ideas
2022-12-08 15:51:31 <veltas      > The ideas are more applicable IMO, and learning Forth will help you understand a bit about how to program
2022-12-08 15:51:38 <veltas      > It's impacted my programming anyway
2022-12-08 15:52:22 <olle        > Of course
2022-12-08 15:53:10 <veltas      > The embedded sphere would be better off if more people knew Forth, even if they didn't use it
2022-12-08 15:56:32 <olle        > I was told forth was best lang out there for eDSL :D
2022-12-08 15:56:51 <veltas      > Probably it is
2022-12-08 15:56:53 <veltas      > That or lisp
2022-12-08 15:57:13 <olle        > Ye, but lips is not that interesting for me actually, dunno why
2022-12-08 15:57:16 <olle        > lisp*
2022-12-08 15:57:37 <olle        > Then you google for "forth embedded dsl" - nothing xD
2022-12-08 15:57:47 <olle        > But I like what I've learned so far
2022-12-08 16:00:42 <veltas      > The 'DSL's in forth are more like libraries really
2022-12-08 16:01:00 <olle        > I still need to learn more
2022-12-08 16:01:05 <olle        > About DOES> etc
2022-12-08 16:01:25 <veltas      > It's been said before on here that Forth isn't really a DSL creator, it can probably do a better job simulating that than a lot of other older languages
2022-12-08 16:01:26 <olle        > And compare with query builder in PHP/Java and SQL eDSL in Forth
2022-12-08 16:01:32 <veltas      > But it's not usually used that way
2022-12-08 16:01:40 <KipIngram   > Yes, the embedded sphere WOULD be better off if more people knew Forth.
2022-12-08 16:01:56 <olle        > would/could/should :)
2022-12-08 16:02:43 <veltas      > olle: Have you considered trying advent of code in Forth?
2022-12-08 16:03:11 <veltas      > The last two years me and some people did the first few problems in forth
2022-12-08 16:03:20 <KipIngram   > DOES> is pretty simple.  You can define a new word foo which itself isused to create new words.  The CREATE part of foo's definition specifies what memory gets allocated for the words foo creates, and the DOES> part specifies what those words (that foo creates) DO when you execute them.
2022-12-08 16:03:23 <veltas      > This year I wanted to focus on other things but I don't mind helping you attempt problems
2022-12-08 16:03:25 <olle        > veltas: I'm going through rosetta code, slowly
2022-12-08 16:03:33 <veltas      > Your choice
2022-12-08 16:03:39 <olle        > Still a noob
2022-12-08 16:03:40 <KipIngram   > : foo  CREATE ...allocate memory... DOES> ...actions... ;
2022-12-08 16:03:43 <KipIngram   > foo bar
2022-12-08 16:03:49 <veltas      > You mean implementing rosetta code?
2022-12-08 16:03:56 <KipIngram   > bar has the memory you allocated, and does the actions when you run it.
2022-12-08 16:03:57 <olle        > veltas: Just reading it so far
2022-12-08 16:04:08 <veltas      > Don't read: do
2022-12-08 16:04:10 <veltas      > Best way to learn
2022-12-08 16:04:11 <olle        > KipIngram: Seems like there are lots of ways to explain the create does> thing :D
2022-12-08 16:04:12 <KipIngram   > The address of bar's memory is first placed on the stack by the system, and then the actions are executed.
2022-12-08 16:04:29 <KipIngram   > Yeah, probably.
2022-12-08 16:04:30 <olle        > It's like monads in haskell haha
2022-12-08 16:04:38 <olle        > One blog post per developer
2022-12-08 16:05:08 <joe9        > if blocks (forth file system) implemented directories, I am curious to know how they would be..
2022-12-08 16:05:22 <veltas      > : CONSTANT CREATE , DOES> @ ;
2022-12-08 16:05:31 <olle        > What's , ?
2022-12-08 16:05:48 <joe9        > dude, read starting forth..
2022-12-08 16:05:51 <dave0       > writes a cell to HERE and increments HERE
2022-12-08 16:05:53 <KipIngram   > BLOCK just doesn't offer that sort of structure.  So "if" it did that, it would "be" any way you wrote it to be - it's nonexistent functionality.
2022-12-08 16:05:54 <olle        > haha trying to
2022-12-08 16:06:01 <veltas      > olle: Append a number to dictionary
2022-12-08 16:06:33 <veltas      > joe9: I have encouraged olle to ask dumb questions in here :P
2022-12-08 16:07:22 <veltas      > joe9: There are 'directories' in blocks, they list ranges of blocks implementing different things
2022-12-08 16:07:23 <olle        > Bigger communities sometimes split beginner channels
2022-12-08 16:07:51 <KipIngram   > That's not "in BLOCK."  It's just part of the data you choose to put into some of your blocks.
2022-12-08 16:08:00 <veltas      > Usually just code the gets directories, there are commands that will also list the first line of many blocks, so the first line usually contains the 'name' of the block or a relevant comment
2022-12-08 16:08:22 <KipIngram   > Of course that's what a directory is in any file system.  It's a file that lists the names and locations of other files.
2022-12-08 16:08:24 <veltas      > There's no one way to do it
2022-12-08 16:09:05 <eris[m]     > id implement it as a string field and a pointer field
2022-12-08 16:09:15 <eris[m]     > this makes it a graph rather than a tree but w/e
2022-12-08 16:09:25 <KipIngram   > I've thought a fair bit about using b-trees to do a file system.
2022-12-08 16:09:31 <KipIngram   > looks like it would work fairly well.
2022-12-08 16:09:31 <olle        > aren't some fs done as trees?
2022-12-08 16:09:37 <KipIngram   > Yes.
2022-12-08 16:09:38 <olle        > HAMMER or HAMMER2?
2022-12-08 16:09:41 <KipIngram   > I'd think most of them.
2022-12-08 16:09:46 <olle        > Right
2022-12-08 16:09:52 <veltas      > All filesystems are trees, what do you think a filesystem structure is? It's a big tree...
2022-12-08 16:10:11 <eris[m]     > Most things we refer to as a 'file system' is a tree
2022-12-08 16:10:23 <eris[m]     > But the unix file system actually isn;t
2022-12-08 16:10:26 <eris[m]     > due to hard links
2022-12-08 16:10:31 <veltas      > If your filesystem structure is just "files are lists of blocks, directories are lists of files" then it is still a tree
2022-12-08 16:10:41 <veltas      > Yes technically eris[m] it's more complicated than a tree
2022-12-08 16:10:44 <veltas      > But you know what I mean
2022-12-08 16:10:46 <KipIngram   > eris: Good point.
2022-12-08 16:11:03 <eris[m]     > Afaik it's a DAG or a full graph
2022-12-08 16:11:10 <eris[m]     > veltas: that assumes you can't create cycles
2022-12-08 16:11:22 <veltas      > I know what a tree is
2022-12-08 16:11:36 <eris[m]     > As soon as a directory can have a directory that's above it in the hierarchy, it stops being a tree
2022-12-08 16:12:17 <veltas      > Then don't allow that, or do if you don't mind it not being a tree
2022-12-08 16:12:52 <eris[m]     > yea
2022-12-08 16:12:56 <olle        > As long as everyone online is disagreeing on semantic details, I'm happy
2022-12-08 16:13:00 <eris[m]     > Lmao
2022-12-08 16:13:16 <veltas      > This is why you can't "link" directories, you can only move, create, or delete empty ones
2022-12-08 16:13:35 <veltas      > I have written a filesystem so I feel a little like I'm being taught to suck eggs
2022-12-08 16:13:47 <olle        > ?
2022-12-08 16:13:50 <olle        > Why?
2022-12-08 16:14:16 <veltas      > Need a more specific question
2022-12-08 16:14:39 <veltas      > Well I don't need it at all, but I am not going to guess at what you're asking
2022-12-08 16:15:05 <olle        > Why write your own fs?
2022-12-08 16:15:22 <veltas      > I wrote it for work, it solved requirements that no open source solution could solve
2022-12-08 16:15:47 <veltas      > Specifically it was designed for a very small space, and optimised for very small files
2022-12-08 16:15:58 <veltas      > i.e. not just padding each file to 512 bytes
2022-12-08 16:16:38 <olle        > Right, embedded
2022-12-08 16:16:40 <olle        > Or?
2022-12-08 16:16:44 <veltas      > Yes
2022-12-08 16:16:54 <olle        > That's like being a real programmer
2022-12-08 16:18:35 <veltas      > And my understanding of Forth influenced the design
2022-12-08 16:18:51 <veltas      > and the implementation
2022-12-08 16:18:58 <veltas      > even though it was written in C
2022-12-08 16:20:27 <olle        > veltas: Subset of C?
2022-12-08 16:28:41 <veltas      > No
2022-12-08 16:28:56 <veltas      > I wrote the filesystem thing in C
2022-12-08 16:37:13 <olle        > Hm
2022-12-08 16:37:20 <olle        > Is that even safe? ^^'
2022-12-08 16:44:30 <veltas      > I write 'safe code' i.e. not horrific to start with, and test a lot, so it's reasonable
2022-12-08 16:45:13 <olle        > Got it :)
2022-12-08 16:45:42 <veltas      > Trust me the bar's not high, I've read source code for firmware that's probably on some computers you own (not my company) and it wouldn't hold up to casual open source scrutiny
2022-12-08 16:45:48 <veltas      > Yet alone a proper security review
2022-12-08 16:50:02 <olle        > I guess
2022-12-08 16:50:17 <olle        > People do medical research on our PHP 4-based application
2022-12-08 16:50:40 <olle        > But age is one of the best betrics for decreased fault density
2022-12-08 16:50:41 <veltas      > Not surprised
2022-12-08 16:50:59 <olle        > It's just not a good metric for changeability :)
2022-12-08 16:51:18 <olle        > But not all domains require changes
2022-12-08 16:51:33 <veltas      > True
2022-12-08 16:51:44 <veltas      > Work with what you have, don't throw the baby out with the bathwater
2022-12-08 16:51:53 <veltas      > But when you can, sure, simplify stuff
2022-12-08 16:52:00 <veltas      > Without ruining your business case in the process
2022-12-08 16:52:40 <gordonjcp   > I look after several hundred mission-critical Windows XP boxes
2022-12-08 16:52:42 <gordonjcp   > <shrug>
2022-12-08 16:53:00 <olle        > XP?
2022-12-08 16:53:03 <gordonjcp   > yup
2022-12-08 16:53:08 <olle        > ooold
2022-12-08 16:53:12 <gordonjcp   > latest version of Windows that this stuff will run on
2022-12-08 16:53:21 <olle        > Hehe
2022-12-08 16:53:38 <gordonjcp   > trying to explain to $new_guy who is complaining that the manual I gave him is 14 years out of date that yes, it's 14 years out of date, and we're running the newest stuff
2022-12-08 16:53:50 <olle        > veltas: We're doing refactoring, but it's a question of budget too. It pays the salaries, I guess.
2022-12-08 16:54:34 <olle        > I'm fighting with our latest senior that no, we can't move everything to Symfony in a reasonable budget
2022-12-08 16:54:46 <veltas      > Well advantage of refactoring is it makes for easier changes in future, you can't predict future requirements
2022-12-08 16:55:08 <olle        > Interface defs "should" be stable.....
2022-12-08 16:55:22 <olle        > But again depends on how the domain changes
2022-12-08 16:55:35 <olle        > Our doesn't change at all, so
2022-12-08 16:55:52 <veltas      > Do you have your own adhoc framework at the moment?
2022-12-08 16:56:14 <olle        > veltas: A team migrated it 80% to Yii during google summer of code :)
2022-12-08 16:56:28 <olle        > We're still working on that, hehe
2022-12-08 16:56:40 <olle        > Really it's a victim of its own success
2022-12-08 16:58:29 <veltas      > Well at least they're both open source frameworks
2022-12-08 17:00:45 <olle        > Sure sure
2022-12-08 17:00:53 <veltas      > If I hear "port this to that framework" I'm really hearing "integrate this 100000 lines of code into your codebase"
2022-12-08 17:01:09 <veltas      > If it's open source at least I can fix it when it inevitably goes wrong
2022-12-08 17:01:52 <olle        > We did a bootstrap conversion this year. We had a senior meeting today where I wanted to discuss the decision that led to the migration.
2022-12-08 17:02:01 <olle        > "I though it would be faster"
2022-12-08 17:02:11 <olle        > Hindsight is 20/20
2022-12-08 17:02:13 <veltas      > Easy to profile this
2022-12-08 17:02:36 <veltas      > Faster to render or faster to design?
2022-12-08 17:02:44 <olle        > Neither
2022-12-08 17:02:53 <olle        > An imagined security issue
2022-12-08 17:03:05 <olle        > And some other stuff. Some kind of group thinking in the end.
2022-12-08 17:03:23 <veltas      > Basically they wanted to use the flashy thing
2022-12-08 17:03:40 <olle        > Or emotional decision making from the CTO. Maybe he panicked from the security report we got from the outside.
2022-12-08 17:03:52 <olle        > It was incorrect in the end.
2022-12-08 17:06:01 <veltas      > It is a constant difficulty figuring out what matters as an engineer
2022-12-08 17:06:16 <olle        > I try to apply risk assessment
2022-12-08 17:06:25 <veltas      > Usually the group will make the wrong decision
2022-12-08 17:06:26 <olle        > But even so, yeah, it's hard
2022-12-08 17:06:48 <olle        > Group psychology is a very specific thing
2022-12-08 17:07:06 <olle        > And tied up to company culture
2022-12-08 17:07:21 <olle        > I think having outsiders look at your shit might be the only way
2022-12-08 17:07:51 <olle        > Like when a company engineer a bridge or tunnel. A completely separate company will also do an analysis on the design. Or so I've heard.
2022-12-08 17:07:57 <veltas      > It's called seeing the wood for the trees
2022-12-08 17:08:50 <olle        > The other day I read in cognitive psychology a technique to analyze your own assumptions.
2022-12-08 17:09:10 <olle        > To write down "IF <I do this or that> THEN <complete the sentence"
2022-12-08 17:09:22 <olle        > Or "IF I AVOID <...> THEN <...>"
2022-12-08 17:09:45 <olle        > But you can't know what you assume!
2022-12-08 17:09:52 <olle        > That's the trap
2022-12-08 17:11:20 <veltas      > You can't analyse everything but there's a state of mind of actually trying to think about things from multiple levels
2022-12-08 17:11:27 <veltas      > And that's better than nothing
2022-12-08 17:11:36 <veltas      > Most people just check out of that kind of thinking
2022-12-08 17:11:54 <KipIngram   > ^^ I agree with this.
2022-12-08 17:12:10 <KipIngram   > Knowing the landscape and "just seeing" that things are going to be ok will get you a long way.
2022-12-08 17:12:42 <KipIngram   > Everything can't be relegated to a bureaucratic process.
2022-12-08 17:13:01 <KipIngram   > But you can't get that message across to corporate types.
2022-12-08 17:13:49 <olle        > Better than nothingm, sure
2022-12-08 17:14:52 <olle        > It's fun to compare with Euclides and his math book that lasted for 1k years
2022-12-08 17:15:13 <olle        > Then in 19th century they noticed he assumed a flat space.
2022-12-08 17:15:24 <veltas      > I think it's worth philosophising about technology generally, think about what fundamentally it is that we're doing, what's fake and what's a fad
2022-12-08 17:16:34 <veltas      > It's the root of me saying stuff like how a web browser is a souped up VT100 etc
2022-12-08 17:17:26 <olle        > "X is just Y"? :)
2022-12-08 17:18:23 <veltas      > Think about that stuff seriously, but also don't just be reductionist: actually test these ideas and be brutally honest about them
2022-12-08 17:19:14 <olle        > At some point a qualitative change happens
2022-12-08 17:19:20 <veltas      > Yes
2022-12-08 17:19:32 <veltas      > The web isn't really just a VT100 but it's sort of descended from it in important ways
2022-12-08 17:19:32 <olle        > And it becomes a fascist radicalization platform lmao
2022-12-08 17:20:13 <veltas      > It helps to understand this family tree sometimes to make it easier to port your skills between different technologies, or to understand how it all comes together
2022-12-08 17:20:40 <olle        > Mm
2022-12-08 17:21:11 <olle        > It's sad considering the potential of comp sci to be a language designer profession
2022-12-08 17:21:19 <olle        > A DSL for each domain
2022-12-08 17:21:30 <olle        > Instead we have these "general" prog lang D:
2022-12-08 17:22:00 <veltas      > I think often you can do almost-DSL stuff just using highly declarative approaches
2022-12-08 17:22:03 <olle        > Where power of expression is limited by design
2022-12-08 17:22:16 <veltas      > And also generally trying to represent the problem with data rather than code
2022-12-08 17:22:26 <olle        > Limited to fit a corporate agenda
2022-12-08 17:22:54 <olle        > There are some strategies, sure, like fluid interfaces described by martin fowler
2022-12-08 17:23:28 <veltas      > I am avoiding the political and societal stuff you're talking about because I don't want to explode the scope of this conversation too much lol
2022-12-08 17:23:38 <olle        > xD
2022-12-08 17:23:54 <olle        > That's for comp sci therapy channel
2022-12-08 17:30:30 <olle        > OK so vt100 does not support css3?
2022-12-08 17:31:03 <eris[m]     > no
2022-12-08 17:32:04 <olle        > well if I can't post my breakfast on it what's the point?
2022-12-08 17:32:06 <olle        > anyway
2022-12-08 17:32:23 <olle        > back to rosetta code
2022-12-08 17:33:02 <veltas      > olle: You can also compare rich text format with HTML and a lot of stuff
2022-12-08 17:33:28 <olle        > "25 nooby Python habits you need to ditch" <--- We need forth youtube vids like these :D
2022-12-08 17:33:59 <olle        > semantic web never happened :/
2022-12-08 17:34:02 <veltas      > I've never watched a video like that