Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[umbraticus]

> you'll need to install hubfs first

not true; just:

srv -c tcp!107.191.50.176!9997 gridchat /n/chat

then read and write from /n/chat/chat in whatever manner you please.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> then read and write from /n/chat/chat in whatever manner you please.

i guess just that `chat` script is located and thus provided by a
hubfs install...

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

Well, I installed all of 9ants (thanks, microftiv), it proved
irresistible, but the Atheros 2414 ethernet adapter isn't supported
and, mysteriously, a 3com 509B isn't detected by the bootstrap
process. It is seen as a PCI device at run time...

And the main goal, really, was TLS/SSH, with quick and slick text
exchanges as a bonus.

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

are there any drivers on other plan9 for your unsupported devices that
we missed pulling in?

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

The Atheros is just not supported, it seems.

It's the 3Com that baffles me. I suspect the two-stage boot (I know
dangerously little about 9front's boot process) somehow fails to pick
up a supported device (3c905b - etherelink3).

Lucio.

PS: It bothers me that I can't serve a 9front workstation from a
9legacy fossil server: the jig-saw puzzle pieces are very subtly
incompatible.

Maybe I ought to remind myself what exactly goes wrong when trying to
netboot a 9front kernel and from there pick a 9front tree to be
served. Naturally, I don't want to have two distinct file servers and
it is way too early for me to feel I should migrate to 9front. I need
to point out that (unless I mis-remember) it is in the Bell Labs
distribution that the gears begin to slip.

On 5/5/19, hiro <23hiro@gmail.com> wrote:
> are there any drivers on other plan9 for your unsupported devices that
> we missed pulling in?
>
>


--
Lucio De Re
2 Piet Retief St
Kestell (Eastern Free State)
9860 South Africa

Ph.: +27 58 653 1433
Cell: +27 83 251 5824
FAX: +27 58 653 1435

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

mycroftiv is using venti+fossil from 9front. i'm sure there is a way :)

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[cinap_lenrek]

> (Incidentally, does 9ants support Go as robustly as 9legacy does? Go
> is critical to my work and so is SSH - sshnet, was it, that provided
> port forwarding? That's another critical feature, so may be OpenVPN,
> but so far that has been too slow on Linux, so it's not a priority.)

recently resurrected sshnet(4). yes, it supports listening connections
now.

openvpn is an absolute kitchensink of options, while still not doing
very much for you.

however, theres a native tinc vpn client (see http://tinc-vpn.org )
i use it to get ipv6 and fixed ipv4 addresses to my home network
from a server, as residential isp's here are unable to offer full
internet service.

--
cinap

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/7/19, cinap_lenrek@felloff.net <cinap_lenrek@felloff.net> wrote:
>
> recently resurrected sshnet(4). yes, it supports listening connections
> now.
>
Thanks, Cinap, nice to hear all that. Did I mention that the rootdir
fix worked like a charm?

And I managed to migrate from 9ants64 to 9ants(32) without losing my
mind. Creating a user other than Glenda took some creativity, but it
worked.

Now, I need to concoct an arbitrator (arbiter) that feeds the right
executables to the 9legacy and 9front systems (in addition to the
architectures - and there's 9atom there too) so I don't have to do it
manually.

Of course, that's wishful thinking, but it's part of a problem that
deserves a solution.

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[cinap_lenrek]

type !rc at the bootargs prompt to get a shell, then run:

grep '^02' '#$/pci/'*ctl

which prints the pci information for all network cards. that
should get us closer to why the probing code fails.

--
cinap

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/5/19, cinap_lenrek@felloff.net <cinap_lenrek@felloff.net> wrote:
> type !rc at the bootargs prompt to get a shell, then run:
>
> grep '^02' '#$/pci/'*ctl
>
> which prints the pci information for all network cards. that
> should get us closer to why the probing code fails.
>
I've been using a handful of those adapters for a long time, the
legacy etherelnk3.c for i386 works flawlessly with them.

I'd rather redirect the effort to document the ethernet adapters that
have been proven to work.  Down here, the challenge isn't antiquated
adapters, it is bad quality ones, that are cheap, but even Linux,
sometimes Windows doesn't support them. And it's all you get.

What I'm suggesting is that we keep the supported hardware list up to
date and provide some details of where it can or could be sourced. It
would certainly make my life a lot simpler (so would moving back to
Europe, but that's a bit less achievable).

That said, I'll see what I can do about the PCI details. I'll probably
have to copy them manually.

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[cinap_lenrek]

> Next plan9.ini's "rootdir". It is no longer documented in
> plan9.ini(8), but the little code that uses it in bootrc is odd as
> well as at odds with the legacy documentation.

yes! we got it all wrong! it all makes sense in the /lib/namespace
what should happen, but bootrc does it wrong.

> My hope was to give it "9front" as a value and have Fossil serve
> that as the root for a diskless 9front workstation, as in

>      bind -b $rootdir /

> Instead, I find "mount -c /srv/boot $rootdir" which is definitely not
> it.

you are absolutely correct. this is wrong. here is the fix:

http://code.9front.org/hg/plan9front/rev/1209e04a3af9

> I'm not sure under what circumstances one target a different
> location for the ROOT directory, I would be interested to hear. Of
> course, this conflict needs to be resolved.

> This is starting to get interesting.

yeah. thank you very much! :-)

> Lucio.

--
cinap

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[cinap_lenrek]

> I have a fun issue where 9front resolution depends on EFI boot method. Via
> firmware interface, I get 1600x900 but via bootloader (EFI file copied to
> esp) I get low resolution.

not so surprising. pure EFI without legacy CSP does not have a VESA BIOS.
all you get is what the bootloader could figure out about the framebuffer
that the firmware set up for us. thats is what we have to work with until
you invoke a native driver to set up a proper mode.

> I can not set resolution via aux/vga so the resolution at boot is the one
> that sticks.

what graphics card is this? when its intel, we have a native driver. it
might just not know about your specific card yet.

> btw: is anyone working on additional WiFi firmware (for example atheros)
> support?

not that i know of.

--
cinap

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Jens Staal]

[-- Attachment #1: Type: text/plain, Size: 985 bytes --]

Den sön 5 maj 2019 16:15 <cinap_lenrek@felloff.net> skrev:

> > I have a fun issue where 9front resolution depends on EFI boot method.
> Via
> > firmware interface, I get 1600x900 but via bootloader (EFI file copied to
> > esp) I get low resolution.
>
> not so surprising. pure EFI without legacy CSP does not have a VESA BIOS.
> all you get is what the bootloader could figure out about the framebuffer
> that the firmware set up for us. thats is what we have to work with until
> you invoke a native driver to set up a proper mode.
>
> > I can not set resolution via aux/vga so the resolution at boot is the one
> > that sticks.
>
> what graphics card is this? when its intel, we have a native driver. it
> might just not know about your specific card yet.
>

http://sysinfo.9front.org/src/269/body

so amd/radeon


> > btw: is anyone working on additional WiFi firmware (for example atheros)
> > support?
>
> not that i know of.
>
> --
> cinap
>
>

[-- Attachment #2: Type: text/html, Size: 1669 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[cinap_lenrek]

> It's the 3Com that baffles me. I suspect the two-stage boot (I know
> dangerously little about 9front's boot process) somehow fails to pick
> up a supported device (3c905b - etherelink3).

is this the amd64 kernel or 386 one? the amd64 one does not include
the etherelnk3 driver. the reason is that the amd64 kernel shares its
drivers with the pc kernel, but not all drivers might be ready for
64 bit pointers. so we only include what we can test and verify.

you can try to add the etherelnk3 line to the pc64 config and
rebuild.

if you have this issue with the 386 kernel, then knowing the pci
device id would be a start.

the boot process is nothing special. we have a bootloader that loads
the kernel. the loader uses BIOS/EFI calls to get the kernel from the
boot media so that it does not need drivers. once the kernel
is taking over, it needs a driver.

--
cinap

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Jens Staal]

[-- Attachment #1: Type: text/plain, Size: 676 bytes --]

Den sön 5 maj 2019 14:34 <cinap_lenrek@felloff.net> skrev:

>
> the boot process is nothing special. we have a bootloader that loads
> the kernel. the loader uses BIOS/EFI calls to get the kernel from the
> boot media so that it does not need drivers. once the kernel
> is taking over, it needs a driver.
>


I have a fun issue where 9front resolution depends on EFI boot method. Via
firmware interface, I get 1600x900 but via bootloader (EFI file copied to
esp) I get low resolution.

I can not set resolution via aux/vga so the resolution at boot is the one
that sticks.

btw: is anyone working on additional WiFi firmware (for example atheros)
support?

[-- Attachment #2: Type: text/html, Size: 1635 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> btw: is anyone working on additional WiFi firmware (for example atheros)
> support?

right now i don't think so.

ath5k and ath9k would likely be welcome, but all newer atheros stuff
seems not worth it (say some important linux wifi driver people, they
complained about non-documented stuff and binay blobs, and are thus
trying to avoid atheros/qualcomm iirc).

i am not aware of any more modern wifi hardware that is both well
documented and available, so unless somebody is willing to do good old
ath5k or ath9k, i'd wait it out a bit more...

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

Gotcha, it is amd64, so I will look at that and report back.

Next plan9.ini's "rootdir". It is no longer documented in
plan9.ini(8), but the little code that uses it in bootrc is odd as
well as at odds with the legacy documentation. My hope was to give it
"9front" as a value and have Fossil serve that as the root for a
diskless 9front workstation, as in

      bind -b $rootdir /

Instead, I find "mount -c /srv/boot $rootdir" which is definitely not
it. I'm not sure under what circumstances one target a different
location for the ROOT directory, I would be interested to hear. Of
course, this conflict needs to be resolved.

This is starting to get interesting.

Lucio.

PS: Thanks a bunch for the support, both you and Hiro.

On 5/5/19, cinap_lenrek@felloff.net <cinap_lenrek@felloff.net> wrote:
>> It's the 3Com that baffles me. I suspect the two-stage boot (I know
>> dangerously little about 9front's boot process) somehow fails to pick
>> up a supported device (3c905b - etherelink3).
>
> is this the amd64 kernel or 386 one? the amd64 one does not include
> the etherelnk3 driver. the reason is that the amd64 kernel shares its
> drivers with the pc kernel, but not all drivers might be ready for
> 64 bit pointers. so we only include what we can test and verify.
>
> you can try to add the etherelnk3 line to the pc64 config and
> rebuild.
>
> if you have this issue with the 386 kernel, then knowing the pci
> device id would be a start.
>
> the boot process is nothing special. we have a bootloader that loads
> the kernel. the loader uses BIOS/EFI calls to get the kernel from the
> boot media so that it does not need drivers. once the kernel
> is taking over, it needs a driver.
>
> --
> cinap
>
>


--
Lucio De Re
2 Piet Retief St
Kestell (Eastern Free State)
9860 South Africa

Ph.: +27 58 653 1433
Cell: +27 83 251 5824
FAX: +27 58 653 1435

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/5/19, cinap_lenrek@felloff.net <cinap_lenrek@felloff.net> wrote:
>
> you can try to add the etherelnk3 line to the pc64 config and
> rebuild.
>
I tried that, but I suspect there is more to it: no errors are
reported on compiling or booting, but the device is still not
identified.

Well, that's not entirely true. I built 9pc64 with elnk3 uncommented,
then discovered the instructions to build 9ants64, so I took that
path. That's where compiling and booting reported no errors (unless a
missing device is considered an error).

When I tried to boot the 9pc64 I built earlier, booting failed. But in
a 9ants context, that may be unrelated. Unfortunately, the diagnostics
that may reveal the nature of the problem flash past far too quickly.
The days of capturing such diagnostics through a serial port seem to
be in the past.

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/5/19, cinap_lenrek@felloff.net <cinap_lenrek@felloff.net> wrote:
>
> [ ... ]
>
> you can try to add the etherelnk3 line to the pc64 config and
> rebuild.
>
Done that, partially, jury is still out. But now I have a test bench
to try things out a little more reliably. Watch this space, I need a
bit of time...

> if you have this issue with the 386 kernel, then knowing the pci
> device id would be a start.
>
We're OK with 386, I downgraded 9ants to 32 bits and it seems to work
pretty well. A good opportunity to benchmark 64-bit versus 32-bit on
the same hardware, something that has got me curious.

(Incidentally, does 9ants support Go as robustly as 9legacy does? Go
is critical to my work and so is SSH - sshnet, was it, that provided
port forwarding? That's another critical feature, so may be OpenVPN,
but so far that has been too slow on Linux, so it's not a priority.)

> the boot process is nothing special. we have a bootloader that loads
> the kernel. the loader uses BIOS/EFI calls to get the kernel from the
> boot media so that it does not need drivers. once the kernel
> is taking over, it needs a driver.
>
Got that sorted, I need to get more familiar with the EFI and GUID
stuff, I know EFI is a trap, but GUID seems an innocent bit of
collateral benefit.

So now I have a somewhat dated instance of 9front on my desktop
courtesy of the VXL Percio diskless workstation, about as limited a
piece of hardware as you can find. That will provide an honest
workbenchin itself. It multiboots various Plan 9 combinations off
compact flash.

That's in addition to a working version of 9ants which will be
promoted to 64-bit once I have a network adapter in it that is
supported by all OSes: Linux Mint, NetBSD and 9ants.

What I need next is definitely Atheros network support, I'm not a
great Wi-Fi fan, but that can't be discounted. I'll start by finding
out what seems to ail the elnk3 driver on amd64. I have half a dozen
3c905[bc] adapters I'd rather not just discard.

Next, a long way off, is bringing 9front, 9ant and 9legacy closer
together, with Minix-3 and NetBSD as even more distant convergences.
Small steps, that is all I can promise. Keep in mind where Minix-3
lurks, before you discount it...

Lucio.

PS: Thanks everyone for your support, I'd like to list you all by
name, but I'm sure to leave someone out. Still, we're a community and
everyone contributes, so my thanks go to all, really, trolls included.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Ethan Gardener]

On Tue, May 7, 2019, at 5:17 AM, Lucio De Re wrote:
>  Keep in mind where Minix-3
> lurks, before you discount it...

I didn't discount Minix-3 until I learned it's recently started using memory protection, and they're having trouble making it work with their IPC.  I'll wait until they've got it sorted.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

Well, when I read Tanenbaum's report from I'm not sure when and all
the regrets when sponsorship opened the doors to contributions, I
assumed that a lesson (not a simple one) was learnt.

I have my own opinions which I prefer to keep to myself, but what is
hard to argue with is that it is all just code. Good code, bad code,
once it works as expected and is not impossible to maintain, one just
needs to throw resources at it (plus management and supervision, those
are resources as well) to make it better.

And that's precisely what I think needs to happen here as well as
anywhere else. My own view of "better" is along the lines of
"smaller", "simpler", but the important thing is to be able to measure
"progress" by some metric that doesn't shift too much.

If Minix-3 fails, in some ways, its lessons will remain. If Linux,
fails, even, the world will move on. Let's just say there shouldn't be
any incentive to push something everyone can learn from to fail...

On 5/8/19, Ethan Gardener <eekee57@fastmail.fm> wrote:
> On Tue, May 7, 2019, at 5:17 AM, Lucio De Re wrote:
>>  Keep in mind where Minix-3
>> lurks, before you discount it...
>
> I didn't discount Minix-3 until I learned it's recently started using memory
> protection, and they're having trouble making it work with their IPC.  I'll
> wait until they've got it sorted.
>
>


--
Lucio De Re
2 Piet Retief St
Kestell (Eastern Free State)
9860 South Africa

Ph.: +27 58 653 1433
Cell: +27 83 251 5824
FAX: +27 58 653 1435

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

feel free to use your own metrics of failure, i don't need any of this
minix stuff

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[umbraticus]

info trav

[9fans] Someone made a Wayland compositor based on Rio, Wio

[Ryan Gonzalez]

[-- Attachment #1: Type: text/plain, Size: 90 bytes --]

https://wio-project.org/

2/10 name, 9/10 demo, can't win it all I guess ¯\_(ツ)_/¯

[-- Attachment #2: Type: text/html, Size: 201 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Rodrigo G. López]

[-- Attachment #1: Type: text/plain, Size: 191 bytes --]

enter the matrix.

On Thu, May 2, 2019, 6:14 AM Ryan Gonzalez <rymg19@gmail.com> wrote:

> https://wio-project.org/
>
> 2/10 name, 9/10 demo, can't win it all I guess ¯\_(ツ)_/¯
>

[-- Attachment #2: Type: text/html, Size: 581 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Skip Tavakkolian]

[-- Attachment #1: Type: text/plain, Size: 242 bytes --]

Can Wio run inside Wio? Broadly speaking, that's the essence of Rio.

On Wed, May 1, 2019, 9:13 PM Ryan Gonzalez <rymg19@gmail.com> wrote:

> https://wio-project.org/
>
> 2/10 name, 9/10 demo, can't win it all I guess ¯\_(ツ)_/¯
>

[-- Attachment #2: Type: text/html, Size: 636 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[David Arnold]

[-- Attachment #1: Type: text/plain, Size: 561 bytes --]

It seems like it runs another Wayland compositor (Cage) in its windows, so the ability to nest compositors is there but I saw no mention of nesting itself. 

No fs yet either although it’s mentioned as a todo. 



d

> On 2 May 2019, at 14:36, Skip Tavakkolian <skip.tavakkolian@gmail.com> wrote:
> 
> Can Wio run inside Wio? Broadly speaking, that's the essence of Rio.
> 
>> On Wed, May 1, 2019, 9:13 PM Ryan Gonzalez <rymg19@gmail.com> wrote:
>> https://wio-project.org/
>> 
>> 2/10 name, 9/10 demo, can't win it all I guess ¯\_(ツ)_/¯ 

[-- Attachment #2: Type: text/html, Size: 1278 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Fazlul Shahriar]

[-- Attachment #1: Type: text/plain, Size: 634 bytes --]

I just gave it a try. Yes, wio can run inside wio.

This is what I had to do to build it:

git clone https://git.sr.ht/~sircmpwn/wio
cd wio
git clone https://github.com/swaywm/wlroots.git subprojects/wlroots
meson build
cd build
ninja
./wio # pops up a window with wio


fhs

On Thu, May 2, 2019 at 1:21 AM Skip Tavakkolian <skip.tavakkolian@gmail.com>
wrote:

> Can Wio run inside Wio? Broadly speaking, that's the essence of Rio.
>
> On Wed, May 1, 2019, 9:13 PM Ryan Gonzalez <rymg19@gmail.com> wrote:
>
>> https://wio-project.org/
>>
>> 2/10 name, 9/10 demo, can't win it all I guess ¯\_(ツ)_/¯
>>
>

[-- Attachment #2: Type: text/html, Size: 2025 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> I just gave it a try. Yes, wio can run inside wio.

i didn't try it and i will not, as wayland is promoted as some
efficient alternative, but is in practice used in a very wasteful way
(so far i thought mostly for distracting 3D animations and scrambling
of information that i'd rather have plain without shadows,
transparency or multiple image transformations or other information
scrambling and concealment tactics.

on the other hand nobody seems to have read the text below the video
(i haven't watched the video) where they indeed advertise the fact
that wio can run inside wio (trivial image transformation
implementation behind, no damage, redraw complications, cause there is
no optimization or such *design* ideas present).

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> Broadly speaking, that's [rio running inside rio] the essence of Rio.

i have lately investigated the code of rio, devdraw and memlayer and
understood a bit the historic connection to bitblt, etc., and i've
been wondering lately where statements like above, which i have heard
before, come from.

they makes no sense to me, that rio nesting feature seems like an
afterthought and involves a lot of back and forth which isn't just
inefficient, but extremely inflexible!

there isn't just a perceived lack of symmetry, direction of
information flow is unordered and opposing the common need.
there are lots of layer violations and no actually transparent
understandable interface.
the documentation is thus also scattered all over those components
that are divided into non-obvious non-coherent layers.
if you knew bitblit it makes more sense, but if you just start (like
me) with modern plan9 proper, consuming the documentation is hard
unless you accidentally stumbled over some descriptions of the
historic progressions, which did seem like a viable excuse for this
clutch that is rio. this old stuff (before rio) must have been the
most latency-sensitive, and bandwidth-constrained piece that had
obvious immediately experienced effects for user interactions. it
makes total sense that bitblit protocol is quite powerful (for the
time) and a lot of stuff is done efficiently in hardware.

i can only assume that the developers knew about the historic debt and
why rio can never be truly elegant as many people (still) advertise,
and that seems to me to be the reason why both sam and rio include
their own incomplete window management that isn't integrated with
anything else in the system in any meaningful way.

other reasons might have been that the system suddenly included
drawterms on windows NT, unixes, macbooks, etc., and inferno, and
nowadays especially plan9port seems to be more important to most 9fans
people than the underlying operating system itself. so why make
another window manager for plan9 if you can make a new theme for
windows95 or ubuntu or whatever instead...

please contest me on my ranty points above, i really would like to see
the real reasons, i'm sure i cannot imagine most of the historic
progression and reasoning. i'm a small kid and i grew up with very
high bandwidth hardware, so... enlighten me please.

and definitely stop this myth about rio being so elegantly nested.
(ask yourself: why does rio need to read the mouse file in order to be
informed about resize events in order to ask for a new window name in
order to attach to the new resize image's window in order to redraw to
the correct size?)

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> and that seems to me to be the reason why both sam and rio include

of course i meant sam and acme ;)
thanks mutantturkey

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Skip Tavakkolian]

[-- Attachment #1: Type: text/plain, Size: 523 bytes --]

One person's epiphany is another's afterthought.

(I've been conditioned by Twitter to not read anything longer than 240
chars. Ditto for responses. I can't speak to points you made later)

On Thu, May 2, 2019, 1:21 PM hiro <23hiro@gmail.com> wrote:

> > Broadly speaking, that's [rio running inside rio] the essence of Rio.
>
> they makes no sense to me, that rio nesting feature seems like an
> afterthought and involves a lot of back and forth which isn't just
> inefficient, but extremely inflexible!
>
>

[-- Attachment #2: Type: text/html, Size: 880 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Dan Cross]

[-- Attachment #1: Type: text/plain, Size: 721 bytes --]

On Thu, May 2, 2019 at 9:24 PM Skip Tavakkolian <skip.tavakkolian@gmail.com>
wrote:

> One person's epiphany is another's afterthought.
>
> (I've been conditioned by Twitter to not read anything longer than 240
> chars. Ditto for responses. I can't speak to points you made later)
>

It wasn't clear to me that there was a question, let alone a point, in that
word salad.

On Thu, May 2, 2019, 1:21 PM hiro <23hiro@gmail.com> wrote:
>
>> > Broadly speaking, that's [rio running inside rio] the essence of Rio.
>>
>> they makes no sense to me, that rio nesting feature seems like an
>> afterthought and involves a lot of back and forth which isn't just
>> inefficient, but extremely inflexible!
>>
>>

[-- Attachment #2: Type: text/html, Size: 1571 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> It wasn't clear to me that there was a question, let alone a point, in that
> word salad.

thank you skip, that was very poetic, entertaining and so enlightening.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

actually i take that back, no thanks to skip, i meant dan!

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

In your defence, Hiro, I quite believe at least some of what you said.
Plan 9 is superior, aesthetically, but not perfect and graphics is no
exception.

Which is unfortunate, because it is also a very difficult field that
attracts mostly those who are prone to take short-cuts and apply
pragmatic solutions and encourages such behaviour even from the most
meticulous of developers.

My own anecdote revolves around font compression. I tried to convert
the tools in the Plan 9 chest (to Go), but I just could not quite pry
that knot loose. I consulted the sources, downloaded the description
of the compression algorithm, you name it. I don't think the
developers intentionally obfuscated the code, I suspect (with
Dijkstra) that the code development lent itself to obfuscation in the
quest for efficiency.

That is a real danger in programming, the real difficulty at the core
of "premature optimisation": the risk is that what gets optimised is
buggy and once reduced to maximal performance, it's the bugs that get
performed maximally and no one can pry them away from the cold
clutches of optimised code.

What this discussion made me think about is something that's been
going through my head a bit of late: given that Plan 9 is so much
simpler than the more popular OSes around, how is it that it does not
outperform them?

Take Skype, for example: I use it exclusively and not exactly
willingly for text conferencing, I hadn't made a voice call in Skype
for years when a recent Skype audio conference turned out not very
successfully.

Now, why isn't there a Plan 9 tool that can beat Skype at at least the
texting portion of its game? Considering how greedy Skype is of
resources, it should not be hard to be (nearly, perhaps even much
more) as comfortable and require fewer resources?

I understand that developers for Plan 9 are few and far between, but
there must be low-hanging fruit out there, or am I also missing some
other really important factor?

On 5/3/19, hiro <23hiro@gmail.com> wrote:
> actually i take that back, no thanks to skip, i meant dan!
>
>


--
Lucio De Re
2 Piet Retief St
Kestell (Eastern Free State)
9860 South Africa

Ph.: +27 58 653 1433
Cell: +27 83 251 5824
FAX: +27 58 653 1435

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

i think you mean the devdraw lempel-ziv compressed images. font images
are just images really (on the memlayer below devdraw), and i also
tried to understand the compression code (the first i ever looked at),
but couldn't.
they did recommend to read that original paper by lempel-ziv and i
didn't - probably i am at fault.

i agree there is inperfection. i find it entertaining to find the
historic reasons how it became like this, and having this excuse
doesn't lessen their work's quality.

i truly think there is some amazing stuff to learn from plan9 design,
but technologic change in the meantime makes a lot of these things
less obvious.
a lot of us younger people have an intuitive understanding only of
newer hardware, and no idea about older bottlenecks, obvious back
then. hence my asking here. trying to check my (i admit somewhat
phantastic) theories :)

> given that Plan 9 is so much simpler than the more popular OSes around, how is it that it does not outperform them?

it does here. disk access on linux is sooooo much higher latency
indeed. it's truly mind-boggling.
my suspicion is that this helps protect the linux disk performance
engineer jobs. we have met such a fellow once during some 9front
benchmarking next to the billiard table. he overheard our conversation
and wanted to offer his services. very nice guy even. probably best
for him that he doesn't know...

> why isn't there a Plan 9 tool that can beat Skype at at least the texting portion of its game

you HAVE to try the hubchat on mycroftiv's grid!

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

also they repeatedly talked about the benefit of starting anew without
unix compatibility: they truly had a good awareness of historic code
cost. i guess i just don't always understand their decision process,
i'm sure it was well thought out!
perhaps sometimes they also did some more pragmatic stuff, and perhaps
don't want to get stuck on talking about that cause there might not be
much to learn from that ;)

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/3/19, hiro <23hiro@gmail.com> wrote:
> i think you mean the devdraw lempel-ziv compressed images. font images
> are just images really (on the memlayer below devdraw), and i also
> tried to understand the compression code (the first i ever looked at),
> but couldn't.
> they did recommend to read that original paper by lempel-ziv and i
> didn't - probably i am at fault.
>
I convinced myself that the algorithm (LZ-77, if I remember right) can
be squeezed into a particular architecture - in this case the C target
language and massaged down to a finely tuned, unreadable mess. My own
failings may have contributed in a big way to that lack of
comprehension.

> i agree there is inperfection. i find it entertaining to find the
> historic reasons how it became like this, and having this excuse
> doesn't lessen their work's quality.
>
That is a valid point, I have a similar concept of DNA and evolution:
making sense of it all is going to be quite a challenge, even though
nothing is absent in the actual implementation. But the details are
layered on top of each other, millions of years of successes and
failures with nothing to flag the one or the other.

> i truly think there is some amazing stuff to learn from plan9 design,
> but technologic change in the meantime makes a lot of these things
> less obvious.
> a lot of us younger people have an intuitive understanding only of
> newer hardware, and no idea about older bottlenecks, obvious back
> then. hence my asking here. trying to check my (i admit somewhat
> phantastic) theories :)
>
There are many things that confuse the hell out of the knowledge I
accumulated from near-as-damn first principles and graphics is high on
that list and has been for a long time. Memory management is pretty
much immediately beneath it. I'm not sure I'll ever properly get a
clear mental model of those complex concepts. Concurrency, I think, is
what makes them difficult for me to grasp.

>> given that Plan 9 is so much simpler than the more popular OSes around,
>> how is it that it does not outperform them?
>
> it does here. disk access on linux is sooooo much higher latency
> indeed. it's truly mind-boggling.

That's nice to hear. The poster child for my gut feeling is acme/Mail.
I used to use it exclusively until various events put my ancient email
address almost entirely out of action. Today, it is Mailman's web
interface that keeps me from switching back to acme/Mail. Having got
Quanstro's upas working (I have sufficient confidence in its
reliability - that's one of Plan 9's less vaunted good features:
software does work) on 9legacy, so now I have to build a mailman
administrator for acme. That isn't going to be quick, even just
starting on it isn't going to happen any time soon.

But the difference even between acme/Mail and mutt was amazing: acme
never lags behind the keyboard or the mouse. You have to lose that for
a while to appreciate it.

> my suspicion is that this helps protect the linux disk performance
> engineer jobs. we have met such a fellow once during some 9front
> benchmarking next to the billiard table. he overheard our conversation
> and wanted to offer his services. very nice guy even. probably best
> for him that he doesn't know...
>
I can believe that. I can even believe that he thinks his role is
justified. But, think about Skype's squiggly line (it is not some
horizontal bubbles - since my upgrade this morning) while someone up
to double the circumference of the earth away types away. Someone
should sue Microsoft for stolen teracycles.

>> why isn't there a Plan 9 tool that can beat Skype at at least the texting
>> portion of its game
>
> you HAVE to try the hubchat on mycroftiv's grid!
>
If you're right (and I see no reason why you shouldn't be, I seem to
remember that mycroftiv's earned his stripes a few times over), it may
be precisely what I need. But can I sell it to a bunch of
semi-literate techies that think Linus Torvalds is the Messiah?

And, please, point me to a useful hubchat starting point? I will
google it in the meantime.

One last thing: the only thing I hold against 9front is the
fragmentation of the Plan 9 camp, not that I think 9front is
responsible, but rather that not enough effort has gone into
converging as often as possible. It's my own project to encourage
convergence just about at all costs. But is it not a one-man exercise,
obviously, specially not this man's :-)

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

> If you're right (and I see no reason why you shouldn't be, I seem to
> remember that mycroftiv's earned his stripes a few times over), it may
> be precisely what I need. But can I sell it to a bunch of
> semi-literate techies that think Linus Torvalds is the Messiah?

if they bother to believe in those gods they can also believe in our gods.

> And, please, point me to a useful hubchat starting point? I will
> google it in the meantime.

http://wiki.9gridchan.org/Hubfs_chat/index.html

you'll need to install hubfs first, from
https://bitbucket.org/mycroftiv/ants9front/src/default/

> One last thing: the only thing I hold against 9front is the
> fragmentation of the Plan 9 camp, not that I think 9front is
> responsible, but rather that not enough effort has gone into
> converging as often as possible. It's my own project to encourage
> convergence just about at all costs. But is it not a one-man exercise,
> obviously, specially not this man's :-)

i agree.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

On 5/4/19, hiro <23hiro@gmail.com> wrote:
>
> if they bother to believe in those gods they can also believe in our gods.
>
There's a problem with believing in any gods: once you do, you're
trapped. It took me fifty years, most of them as an atheist, to feel
certain that there is no afterlife.

I still realise that I could never attain a state in which religion
plays absolutely no role, yet my companion, having been brought up
without the god delusion, lives daily in such a state. So, it is
possible, but not attainable.

Lucio.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

that's not my problem lucio. my deal with glenda has been made and i
am promised great low latency 9p fileservers in my afternamespaces.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Lucio De Re]

Oh! A pact with Tux is far better for instant gratification ;-).


On 5/5/19, hiro <23hiro@gmail.com> wrote:
> that's not my problem lucio. my deal with glenda has been made and i
> am promised great low latency 9p fileservers in my afternamespaces.
>
>


--
Lucio De Re
2 Piet Retief St
Kestell (Eastern Free State)
9860 South Africa

Ph.: +27 58 653 1433
Cell: +27 83 251 5824
FAX: +27 58 653 1435

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Ethan Gardener]

On Fri, May 3, 2019, at 1:34 PM, hiro wrote:
> a lot of us younger people have an intuitive understanding only of
> newer hardware, and no idea about older bottlenecks, obvious back
> then.

Meanwhile, I have an intuitive understanding of older hardware and no idea about newer bottlenecks.  And I'm starting to develop an OS.  Great.  Planning is going insanely well; today I thought I could get the filesystem to 'fall out of' variable allocation.  (There was a bit more to it than that. ;)

I can't really help with your actual question, hiro, other than to say, "gotta go fast."  Compression is one thing I've never really looked into beyond RLE.


On the actual thread topic, I guess Wio is cool.  If it works as well as I think it does, I shall have to improve my opinion of Wayland.

Stray thought:  Why didn't the suckless folk ever use Xnest instead of mucking about with things which require application support?  Not that Xnest has worked 100% for years, but they could have tried to fix it.  I mentioned it to them but they were already hyped up about tabbed.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Jens Staal]

[-- Attachment #1: Type: text/plain, Size: 1064 bytes --]

>
>
> On the actual thread topic, I guess Wio is cool.  If it works as well as I
> think it does, I shall have to improve my opinion of Wayland.
>

The developer of wio is the same that wrote sway (an i3-like Wayland
compositor) and wlroots, which is now used by many projects. Sway is
definitely a really nice environment to work in.


> Stray thought:  Why didn't the suckless folk ever use Xnest instead of
> mucking about with things which require application support?  Not that
> Xnest has worked 100% for years, but they could have tried to fix it.  I
> mentioned it to them but they were already hyped up about tabbed.
>

I would say that I prefer a tabbed+tiled environment and would love a Rio
variant that behaves like i3 or DWM with workspaces and tiling (dmenu does
not make sense on Plan9 due to namespaces - so execution from terminal
windows like in Rio would still be the same).

If I am making a "laundry list" of wishes I would also love such an
interface to be based on the solarized colour schemes (dark, light).


>

[-- Attachment #2: Type: text/html, Size: 1796 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[hiro]

archlinux user detected.

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[岡本健二]

[-- Attachment #1: Type: text/plain, Size: 856 bytes --]

I made wio on ubuntu 18.04, and confirmed wio can run nested.
It looks like Plan 9's rio!
To compile wio, especially wiroots on 18.04, we have to make meson wlroots
cage alacritty wayland-protocols for more recent versions.

Kenji

2019年5月3日(金) 9:49 Skip Tavakkolian <skip.tavakkolian@gmail.com>:

> One person's epiphany is another's afterthought.
>
> (I've been conditioned by Twitter to not read anything longer than 240
> chars. Ditto for responses. I can't speak to points you made later)
>
> On Thu, May 2, 2019, 1:21 PM hiro <23hiro@gmail.com> wrote:
>
>> > Broadly speaking, that's [rio running inside rio] the essence of Rio.
>>
>> they makes no sense to me, that rio nesting feature seems like an
>> afterthought and involves a lot of back and forth which isn't just
>> inefficient, but extremely inflexible!
>>
>>

[-- Attachment #2: Type: text/html, Size: 1569 bytes --]

Re: [9fans] Someone made a Wayland compositor based on Rio, Wio

[Bakul Shah]

On May 2, 2019, at 4:10 AM, hiro <23hiro@gmail.com> wrote:
>
>> Broadly speaking, that's [rio running inside rio] the essence of Rio.
>
> i have lately investigated the code of rio, devdraw and memlayer and
> understood a bit the historic connection to bitblt, etc., and i've
> been wondering lately where statements like above, which i have heard
> before, come from.
>
> they makes no sense to me, that rio nesting feature seems like an
> afterthought and involves a lot of back and forth which isn't just
> inefficient, but extremely inflexible!

Look at the moon, not the pointing finger.

[Answer kept below Skip's Twitter read limit.]