Discussing a new design of partially Squeak-based OS
Locked
Discussing a new design of partially Squeak-based OS
 
|
Hello folks. New member here. Forgive me if this is the wrong channel.
I did a talk at the FOSDEM conference a little under 2 weeks ago about a proposal for a new type of OS, based entirely on FOSS code. A major component of it would be Squeak. I was wondering if this would be an appropriate place to talk about it, but I don't want to spam you with links if it's the wrong channel. If this is not the right place, then where might be? -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Re: Discussing a new design of partially Squeak-based OS
|
|
Liam,
> Hello folks. New member here. Forgive me if this is the wrong channel. > > I did a talk at the FOSDEM conference a little under 2 weeks ago about > a proposal for a new type of OS, based entirely on FOSS code. A major > component of it would be Squeak. I was wondering if this would be an > appropriate place to talk about it, but I don't want to spam you with > links if it's the wrong channel. I post the link for you, then: https://fosdem.org/2021/schedule/event/new_type_of_computer/ > If this is not the right place, then where might be? There are much better places to have this discussion, but they are all pretty much dead. So I would not have any objection to having it here. Let's see what other people think. http://mailman.mit.edu/mailman/listinfo/fonclist http://lists.tunes.org/mailman/listinfo -- Jecel |
Re: Discussing a new design of partially Squeak-based OS
|
|
On Fri, 19 Feb 2021 at 18:45, Jecel Assumpcao Jr <[hidden email]> wrote:
> > I post the link for you, then: > > https://fosdem.org/2021/schedule/event/new_type_of_computer/ Hi again, Jecel, and thank you. :-) Since we conversed on Hacker News, I have downloaded and installed Squeak and am working my way through the Red, Green and Blue Books. I may be some time, though. I have put the slides, the script and a downloadable video on my blog as well, here: https://liam-on-linux.livejournal.com/77065.html > There are much better places to have this discussion, but they are all > pretty much dead. So I would not have any objection to having it here. > Let's see what other people think. Sounds good to me. -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Jecel Assumpcao Jr
Hi Guys,
Is there any of Intel's 3D Xpoint as byte addressable ram or is it just SSDs? Lou On Fri, 19 Feb 2021 14:45:41 -0300, "Jecel Assumpcao Jr" <[hidden email]> wrote: >Liam, > >> Hello folks. New member here. Forgive me if this is the wrong channel. >> >> I did a talk at the FOSDEM conference a little under 2 weeks ago about >> a proposal for a new type of OS, based entirely on FOSS code. A major >> component of it would be Squeak. I was wondering if this would be an >> appropriate place to talk about it, but I don't want to spam you with >> links if it's the wrong channel. > >I post the link for you, then: > >https://fosdem.org/2021/schedule/event/new_type_of_computer/ > >> If this is not the right place, then where might be? > >There are much better places to have this discussion, but they are all >pretty much dead. So I would not have any objection to having it here. >Let's see what other people think. > >http://mailman.mit.edu/mailman/listinfo/fonclist > >http://lists.tunes.org/mailman/listinfo > >-- Jecel > Louis LaBrunda Keystone Software Corp. SkypeMe callto://PhotonDemon |
Re: Discussing a new design of partially Squeak-based OS
|
|
On Fri, 19 Feb 2021 at 20:58, Louis LaBrunda <[hidden email]> wrote:
> > Hi Guys, > > Is there any of Intel's 3D Xpoint as byte addressable ram or is it just SSDs? 3D Xpoint is all byte-addressable and byte-erasable, yes. It is also available in DIMM form, called NVDIMMs, that plug directly into the machine's memory bus and make it appear in the memory map. Linux being Linux, however, the primary usage model is to partition it, format it, and use it as a very very fast disk drive. :-( I wrote the section of SUSE Linux Enterprise Server's documentation concerning NVDIMMs. :-) -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Re: Discussing a new design of partially Squeak-based OS
|
|
Hi Liam, I saw your talk and enjoyed it very much. I think people on this chain will have a lot to add, particularly about projects like SqueakNOS. I want to point you to the most recent of these kinds of efforts that I know about called CogNOS (https://github.com/nopsys/CogNOS). I am not affiliated with this project though and don't really know the status of it. On Fri, Feb 19, 2021 at 3:10 PM Liam Proven <[hidden email]> wrote: I did a talk at the FOSDEM conference a little under 2 weeks ago about -- Eric |
Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Liam Proven
Hi Guys,
Thanks for the reply. >> Is there any of Intel's 3D Xpoint as byte addressable ram or is it just SSDs? >3D Xpoint is all byte-addressable and byte-erasable, yes. >It is also available in DIMM form, called NVDIMMs, that plug directly >into the machine's memory bus and make it appear in the memory map. >Linux being Linux, however, the primary usage model is to partition >it, format it, and use it as a very very fast disk drive. :-( When Intel first announced this memory (years ago now) I realized right away it's potential as an object database for Smalltalk. The OS would need to be changed to assign it to a given userId or program. The Smalltalk VM would need to recognize it as a special object memory that contained persistent objects and deal with it properly. I don't think the change to the OS or the VM should be a very big deal. I hadn't heard much about the memory a while. Searching now I see that 256GB goes for about $1900 and 512GB for $9000, not cheap. If the prices come down and the size goes up, it would be a great object database with all the ease of accessing the data (objects) via Smalltalk. Lou >I wrote the section of SUSE Linux Enterprise Server's documentation >concerning NVDIMMs. :-) -- Louis LaBrunda Keystone Software Corp. SkypeMe callto://PhotonDemon |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by darth-cheney
On Fri, 19 Feb 2021 at 22:02, Eric Gade <[hidden email]> wrote:
> > Hi Liam, > > I saw your talk and enjoyed it very much. Thank you -- I'm very pleased and flattered to hear that! > I think people on this chain will have a lot to add, I rather thought they might. ;-) > particularly about projects like SqueakNOS. I am aware of it and have read about it in as much depth as I was able to follow, some time ago. I made a very brief mention of SqueakNOS in the talk when introducing Squeak itself, when I said: « The latest development in Smalltalk is an educational programming environment called Squeak. Its runtime is implemented in C, *and there’s even a bare-metal version*, but Smalltalk runtimes have been built in other languages, including Javascript. The JIT technology in Oracle’s JVM started off as a Smalltalk engine called StrongTalk. » > I want to point you to the most recent of these kinds of efforts that I know about called CogNOS (https://github.com/nopsys/CogNOS). > I am not affiliated with this project though and don't really know the status of it. That flavour is new to me and I was previously unfamiliar with it. Thanks for the info. I have had a quick read and I am glad to see that there is some life, but no commits in 2 years suggests perhaps not that much... I was taken aback to see commits by "charig" but it is not Francis Charig as for a wild moment I thought it might be. Charig was one of the founders of Tao Group, who created Taos and later Elate and Intent, perhaps the most technically impressive OSes I've ever seen. https://en.wikipedia.org/wiki/Francis_Charig But this ChariG is Guido Chari and the username seems to be a coincidence. I very much do not want to tread on anyone's toes or upset anyone, but let me try to explain why I proposed something different to the SqueakNOS approach. I explain my reasoning in more detail in the talk itself -- the script is here, BTW: https://docs.google.com/document/d/1wM1-c7euvQaRaCL4hKCaE8VRJ8OxpVdkzHxQF_8X5So/edit But I had two main goals: [1] eliminate absolutely as much of the existing legacy technology stack as possible. This very much includes eliminating the C language. [2] that my proposed design should be portable and readily and easily adapted to different CPU architectures; at a minimum, x86-64 and ARM. As I understand it -- please correct me if I am wrong -- SqueakNOS is C-based and only supports x86. A third goal, one perhaps much more difficult to achieve, was to offer choices, and as a second option, harder to implement and based on less-complete code, I also proposed, instead of a Smalltalk layer, a Lisp layer, to include both a conventional Lisp and the friendlier, higher-level Dylan language. In an ideal world, perhaps these could be to some degree interoperable. A man can dream. Such goals imply, to me, a relatively mature, high-performance, small OS underneath, delivering some degree of portability -- something written in a type-safe, memory-managed language, with multiprocessor support and networking and so on already present, so that these things do not have to be implemented in the higher-level system. My suggestion for that OS is the Oberon system, specifically the version called A2. It has its own windowing GUI, called Bluebottle, which I think would be largely irrelevant to Squeak, but it already has a native x86 version (alongside Windows, Linux and macOS hosted forms) and formerly also supported StrongARM and a number of other CPUs. It is SMP- and multicore-capable, supports Ethernet and TCP/IP, a USB stack and so on. -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Louis LaBrunda
On Fri, 19 Feb 2021 at 22:46, Louis LaBrunda <[hidden email]> wrote:
> > I hadn't heard much about the memory a while. Searching now I see that 256GB goes for about $1900 and 512GB for $9000, > not cheap. If the prices come down and the size goes up, it would be a great object database with all the ease of > accessing the data (objects) via Smalltalk. I don't know where you are looking but those figures look *wildly* off to me. A tiny 16GB disk is $20: https://www.amazon.com/Intel-Optane-Memory-Module-MEMPEK1W016GA/dp/B07CGYTXQK/ A bigger 256GB one only $99: https://www.amazon.com/Intel-760P-256GB-80mm-PCIe/dp/B078VBL3T9/ Intel also offers NVDIMM modules but those do not currently appear on Amazon. https://www.extremetech.com/computing/259155-micron-launches-new-nand-based-dimms-intel-announces-optane-dimms https://techunplugged.io/2020/07/03/intel-announces-2nd-gen-optane-persistent-memory-modules/ -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Liam Proven
Liam,
> As I understand it -- please correct me if I am wrong -- SqueakNOS is > C-based and only supports x86. CogNOS is a bit more dependent on C than SqueakNOS, but neither is really built on C. The bulk of the code is written in Slang, which is a restricted subset of Smalltalk-80. This gets translated to C which is then compiled to whatever processor you use but it would be a relatively small project to eliminate the C step. If you want to use some of the VM extensions that happen to be written in C instead of Slang then you would indeed need C. The drivers are written in normal Smalltalk-80 and so don't depend at all on the x86 (or C). But they do mean you can use only a particular Ethernet card or a specific interrupt chip. You need PS/2 keyboard and mouse. So I would say it only supports PCs, but the processor can be any one you like. That is for the interpreter. If you want to use the Cog compiler then the processor must be ARM32, ARM64, x86, AMD64 or MIPS since those are currently the ones for which code generators have been implemented. > My suggestion for that OS is the Oberon system, specifically the > version called A2. [...] In the early 1990s a friend saw that I was designing Smalltalk computers and suggested that I should consider Oberon. I was already familiar with the older Lilith project and was happy to try it on the 386 machine I had at the time. While I liked it a lot I felt it was too static to grow very far. I felt the same thing about the C++ based BeOS and though they had some great results and amazing applications I don't think I was wrong about that. -- Jecel |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Liam Proven
Hi Liam,
On Feb 19, 2021, at 11:41 AM, Liam Proven <[hidden email]> wrote:
I’m the architect of Cog. I’d be interested in discussing where I hope Cog is going technologically to try abd find a direction that suits both your goals and the various goals of people working on Cog.
|
Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Liam Proven
Hi Liam,
On Sat, 20 Feb 2021 00:58:21 +0100, Liam Proven <[hidden email]> wrote: >On Fri, 19 Feb 2021 at 22:46, Louis LaBrunda <[hidden email]> wrote: >> >> I hadn't heard much about the memory a while. Searching now I see that 256GB goes for about $1900 and 512GB for $9000, >> not cheap. If the prices come down and the size goes up, it would be a great object database with all the ease of >> accessing the data (objects) via Smalltalk. > >I don't know where you are looking but those figures look *wildly* off to me. I was talking about RAM cards (NVDIMM modules) not SSDs: https://www.newegg.com/p/0RN-0020-000M5?item=0RN-0020-000M5&source=region&nm_mc=knc-googleadwords-pc&cm_mmc=knc-googleadwords-pc-_-pla-_-memory+%28desktop+memory%29-_-0RN-0020-000M5&gclid=CjwKCAiAg8OBBhA8EiwAlKw3kt54KfX2xMGJSIgTUKV4qnQ3LQp8ABciwn7MNm_aMh3FPmV6LN_Q_xoCvG4QAvD_BwE&gclsrc=aw.ds https://www.newegg.com/intel-512gb-ddr-t/p/0RN-0020-00082 http://www.nextwarehouse.com/item/?4037146_g10e&gclid=CjwKCAiAg8OBBhA8EiwAlKw3khpC5DI7GaeCiwG8JA85tMWgSoOBiY66kGQ3N9UXErZulLmrrPkHvRoCA94QAvD_BwE The idea is that the objects placed in this area of ram are permanent (safe stored) and therefor constitute a database. The Smalltalk programmer need only indicate that an object be part of the database, the VM then puts it (and other objects it points to) in this special area of memory. This requires some changed to the OS and VM to police ownership of the database but I wouldn't think that would be too complicated. As for the price differences between the RAM and SSD cards, my guess (I don't know for sure) is that even though the memory can be byte addressable, the SSD cards are configured as 512 byte blocks that must be read or written as blocks. That probably brings the price way down. Lou >A tiny 16GB disk is $20: > >https://www.amazon.com/Intel-Optane-Memory-Module-MEMPEK1W016GA/dp/B07CGYTXQK/ > >A bigger 256GB one only $99: > >https://www.amazon.com/Intel-760P-256GB-80mm-PCIe/dp/B078VBL3T9/ > >Intel also offers NVDIMM modules but those do not currently appear on Amazon. > >https://www.extremetech.com/computing/259155-micron-launches-new-nand-based-dimms-intel-announces-optane-dimms > >https://techunplugged.io/2020/07/03/intel-announces-2nd-gen-optane-persistent-memory-modules/ Louis LaBrunda Keystone Software Corp. SkypeMe callto://PhotonDemon |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Eliot Miranda-2
On Sat, 20 Feb 2021 at 15:38, Eliot Miranda <[hidden email]> wrote:
> > Hi Liam, > > A good shortcut is to read Squeak by Example, which has been updated to refer to current Squeak here: https://github.com/hpi-swa-lab/SqueakByExample-english/releases/download/5.3/SqueakByExample_5_3.pdf Thanks! Yes, I downloaded that too, but a former-Smalltalker friend has been trying to get me to read the R/G/B books for a decade now... > I’m the architect of Cog. I’d be interested in discussing where I hope Cog is going technologically to try abd find a direction that suits both your goals and the various goals of people working on Cog. By all means! If enough people seem to be interested, I could start a new mailing list just for it... This one does have its own agenda, I think, no? -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Re: Discussing a new design of partially Squeak-based OS
|
|
In reply to this post by Louis LaBrunda
On Sat, 20 Feb 2021 at 18:17, Louis LaBrunda <[hidden email]> wrote:
> > I was talking about RAM cards (NVDIMM modules) not SSDs: Aha! > https://www.newegg.com/p/0RN-0020-000M5?item=0RN-0020-000M5&source=region&nm_mc=knc-googleadwords-pc&cm_mmc=knc-googleadwords-pc-_-pla-_-memory+%28desktop+memory%29-_-0RN-0020-000M5&gclid=CjwKCAiAg8OBBhA8EiwAlKw3kt54KfX2xMGJSIgTUKV4qnQ3LQp8ABciwn7MNm_aMh3FPmV6LN_Q_xoCvG4QAvD_BwE&gclsrc=aw.ds > > https://www.newegg.com/intel-512gb-ddr-t/p/0RN-0020-00082 > > http://www.nextwarehouse.com/item/?4037146_g10e&gclid=CjwKCAiAg8OBBhA8EiwAlKw3khpC5DI7GaeCiwG8JA85tMWgSoOBiY66kGQ3N9UXErZulLmrrPkHvRoCA94QAvD_BwE *Ouch*. Still, it's early days. I also think it might be more relevant to compare to RAM prices than flash SSD prices. E.g. $3.5K for a 256GB DDR4 DIMM: https://www.memorycow.co.uk/256gb-ddr4-pc4-23400-2933mhz-288-pin-dimm-ecc-lrdimm-memory-ram > The idea is that the objects placed in this area of ram are permanent (safe stored) and therefor constitute a database. > The Smalltalk programmer need only indicate that an object be part of the database, the VM then puts it (and other > objects it points to) in this special area of memory. This requires some changed to the OS and VM to police ownership > of the database but I wouldn't think that would be too complicated. Seems reasonable. > As for the price differences between the RAM and SSD cards, my guess (I don't know for sure) is that even though the > memory can be byte addressable, the SSD cards are configured as 512 byte blocks that must be read or written as blocks. > That probably brings the price way down. There is a talk about another PMEM-centric OS on Youtube, Twizzler, that mentions this in passing. https://youtu.be/bSNda9EzNOI It's quite long (~1h20m) but so bursting with ideas that it's a packed hour+ even at 1.25x speed, which necessitated pausing and rewinding a few times for me. -- Liam Proven – Profile: https://about.me/liamproven Email: [hidden email] – gMail/gTalk/gHangouts: [hidden email] Twitter/Facebook/LinkedIn/Flickr: lproven – Skype: liamproven UK: +44 7939-087884 – ČR (+ WhatsApp/Telegram/Signal): +420 702 829 053 |
Discussing a new design of partially Squeak-based OS
|
|
Hi Liam,
snip... >*Ouch*. > >Still, it's early days. I also think it might be more relevant to compare to RAM prices than flash SSD prices. I agree. If the object database can fit in the 512GB (and the OS and Smalltalk VM are changed to do what is needed) $9,000 isn't that bad. Being able to use Smalltalk collections in an object that would ordinarily be an SQL table row and require complex SQL tables and code, makes the cost of the memory look cheep. Lou -- Louis LaBrunda Keystone Software Corp. SkypeMe callto://PhotonDemon |
Re: Discussing a new design of partially Squeak-based OS
|
|
[_]— is there a Squeak option to have Squeak do an Oberon like infinite size desktop plus Oberon like zooming ? i don’t think so but if not wouldn’t that be easy to configure ? To make Squeak Roassal like ? with the Connectors etc <Cntrl> plus scroll up = zoom out <Cntrl> plus scroll down = zoom in and also mouse gestures some quick gesture plus a drag up zooms out drag down=in i will add Oberon Button words to my thing as anything that is underlined is a Button so no Smalltalk selecting first anymore and if the cursor is Arrow keyed into the underline the underline gets selected and if you then hit <Enter> it tries to run it . If you hit <Enter> twice within a time limit then the selection gets replaced by a CrLf . if you do this in helpMode: it tells what is happening with links . And then i hook up to Roassal to get the infinite zoomable desktop thing going but the keyboard control of the zooming has to be quicker maybe it is now i d k . An Oberon like Squeak desktop option would be good in my opinion and the Oberon Button word is easy too it’s just removing the selecting part as desired by underlining in my case . A popup to lay an icon over the word could be good . In a more generalized thing you select some Smalltalk popup [ Make A Link ] and replace it with a hyperlink or a graphic Button but i like the simple Oberon way too. i believe this has already been done before for web links in Squeak or VisualWorks Thank you for reminding me of these Oberon ideas Liam i will implement them at once into a #todo who knows when it will get done just in time I saw a JavaScript thing that had Lisp under Smalltalk but i forget everything . Amber ? It was looking pretty good a long time ago but i only saw it work once i can’t remember the name if it’s not Amber i wish i could it was looking pretty good What i want is a Lisp dialect that is highly isomorphic to Smalltalk . I have one . ( at:ifAbsentPut: aDictionary key [...] )<——[ a SmACC pre processing could do this trick ] or ( at:ifAbsentPut: aDictionary at:key ifAbsentPut:[...] )<—-[ the extra keywords are comments ] ( at:ifAbsentPut: aDictionary atKey: key getValue-ifAbsentPutNewValue:[...] )<—-[ is the same message send but with more verbose comments ] [ x ]— i made a demo in 2011 in Vegas including a very simple idea for a Debugger feature that steps into a Macro call like one steps into a Message send i suppose it could optionally step the String expansion process too and then step the result it is very easy to do this you just have to say ok i do it maybe it’s already done Package name? [_]— You can do Macros in Smalltalk easily just mark a Method that returns a String as a Macro which can be done in Dolphin by defining a dynamic Category of which there are two pre existing examples Public and Private so just add Macro to those two now why didn’t i think of that I want this A2 you speak of i think i tried to download Oberon a while ago but the page was a ghost town run around i felt like maybe i did get it but i never saw it run so if there is something more slick and shining i want it There is a LispOS called mezzanine or something on GitHub OS lite maybe So what you are saying really is that you want a port of Squeak to A2 is this not right ? of course it is a modified A2 that is not disk centric On Sun, Feb 21, 2021 at 09:03 Louis LaBrunda <[hidden email]> wrote: Hi Liam, |
Re: Discussing a new design of partially Squeak-based OS
|
|
[_]— and then if you use custom colors on the code you can hide the initial at:ifAbsentPut: by coloring it cyan 255 and setting the Font to Arial 8 if you want it to be more Smalltalk looking Smalltalk is the most self documenting language i think On Sun, Feb 21, 2021 at 12:00 Kjell Godo <[hidden email]> wrote:
|
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |