<felix_>
having a high quality bsd licensed ip stack would be awesome
<felix_>
whitequark: does it also have ipv6 support?
<rqou>
iirc lwip does, and azonenberg found a CVE in it :P
<whitequark>
er, I don't remember. I think so?
digshadow1 has joined ##openfpga
digshadow has quit [Ping timeout: 260 seconds]
<rqou>
oh nvm, azonenberg's exploit was in microchip's custom stack
m_w has quit [Quit: leaving]
<cr1901_modern>
Is "high quality" and "small tcp/ip stack" mutually exclusive?
<rqou>
vxworks? :P
<rqou>
if you include "open source" then almost certainly
<cr1901_modern>
I am highly unlikely to even consider an non-FOSS solution for anything in 2016, *IF* I can help it.
<rqou>
but vxworks is so configurable! /s
<rqou>
my very limited interaction with vxworks gave me a first impression of "yeah, it's basically a less-ram-hungry linux"
<whitequark>
rqou: what do you think of han unification
<rqou>
ask @eevee :P
<whitequark>
i wondered if you had some particular insight
<rqou>
imho the PRC seems to have gotten much fewer problems out of it compared to e.g. Japan
<whitequark>
there's a spin that american liberals fancy, which is that it was a form of cultural imperialism or... something
<whitequark>
i looked up the meeting minutes of the committee which decided it
<whitequark>
it had one person with a western name, in a largely ceremonial role, out of several dozen
<rqou>
it seems that after moving to simplified characters people in the PRC had fewer complaints about minor style differences like e.g. "number of strokes in the 'grass' radical"
<rqou>
i don't really agree with that spin
<rqou>
at best I would agree with "China bullied the rest of east asia"
<whitequark>
I see
<rqou>
overall i think it was a dumb idea
<rqou>
they didn't actually succeed in fitting everything in 2^16 in the end anyways
<rqou>
on a different note, the chinese GB 18030 standard requires astral plane support
<rqou>
it's also a superset of unicode and totally a valid UTF
<azonenberg>
cr1901_modern: lol apparently
<azonenberg>
i'm working on making a solid tcp/ip stack for antikernel
<azonenberg>
but it's all hardware offload and not (yet) optimized for size
<azonenberg>
There are also some features, like packet fragmentation, that i do not ever intend to support, by design
<azonenberg>
The subset of RFC features that are implemented, however, should be solid
<rqou>
but what will you do when comcast's broken dhcp sends you an mtu of 576? :P
<cr1901_modern>
azonenberg: Okay, but me being able to use your tcp/ip stack meaningfully is years off :)
<azonenberg>
rqou: Set TCP max segment size to 576 - header size
<cr1901_modern>
And I don't actually know how TCP/IP works; I learned more than I would've liked when I set up a mail server and had to use Wireshark to debug an ARP problem (please I beg you don't ask)
<azonenberg>
That's what I did in my home LAN years ago when their PPPoE cut 8 bytes off my MTU
<rqou>
and then what happens when your housemate tries to play titanfall which sets don't fragment flags?
<rqou>
:P
<rqou>
(yes, this happened)
<azonenberg>
rqou: Most linux code/stacks set DF flags
<azonenberg>
iirc if you just use default linux socks you get DF on all traffic
<azonenberg>
I had problemds with web pages randomly failing to load
<azonenberg>
if they sent a TCP segment between the set MTU and the actual MTU
<rqou>
somehow titanfall had problems with that
<rqou>
the fix was to ignore the mtu from comcast and always make it 1500
<azonenberg>
oh, this was different
<azonenberg>
this wasnt a DHCP MTU
<azonenberg>
this was an actual hard MTU on the link
<azonenberg>
like, the DLSAM only took 1500 byte total packet size, including the PPPoE header
<rqou>
fortunately i never had to use pppoe and never had this problem
<azonenberg>
So your MTU had to be 1492 before encapsulation
<azonenberg>
Which meant the TCP MSS was 1492 - TCP/IP header size
<azonenberg>
not 1500 - headers
<rqou>
yeah, iirc iptables has a mssfix hack thing for this
<azonenberg>
i did it on the routr
<azonenberg>
dont remember specifics
<rqou>
the iptables hack is intended to be used on the router
<rqou>
overall does PMTU discovery actually work yet?
<rqou>
can I run my gigabit lan with 9000 jumbo frames but have the internet connection still be 1500?
<azonenberg>
No idea
<azonenberg>
most of my FPGA stuff is 1500 only
<azonenberg>
b/c a block RAM is 2048
<rqou>
make the mtu 2048? :P
<azonenberg>
so i could easily support slight jumbo
<azonenberg>
But i never bothered yet
<azonenberg>
My shorter term priority is making a full ipv6 transition
<azonenberg>
I want to stop using ipv4 for all non-internet-facing traffic
<azonenberg>
iow, any vlan that does not route to the internet should not have ipv4 addressing, at all
<azonenberg>
and go 100% SLAAC
<cr1901_modern>
I have too many vintage machines to play with at home to remove ipv4. And don't see myself using Teredo or whatever as a stopgap solution.
<azonenberg>
I'm going to keep ipv4 on one management network for some legacy gear that doesnt do v6
<azonenberg>
and on anything internet-facing
<azonenberg>
everything else, pure v6
<cr1901_modern>
I've been thinking of making an accelerator board for crypto operations to implement SSL on my vintage machines as well. Much more satisfying then "attaching everything to an RPi and making it do all the work" :P
<Lord_Nightmare>
well, there was an LZS STAC compression card for nubus back in the day
<Lord_Nightmare>
a special version of stac for mac could use it for drive compression
<rqou>
my father claims he knew some of the STAC people and helped them do something or other
<Lord_Nightmare>
wait, was your dad the one who designed that card?
<rqou>
definitely no
<Lord_Nightmare>
i think balrog has one
<rqou>
he helped them find a patent or something
<balrog>
the card has an ASIC that implements LZSS compression
<Lord_Nightmare>
find or file? iirc the stac lzs patent was proven invalid at some point
<balrog>
err not LZSS, LZS
<Lord_Nightmare>
not until after they sued microsoft and got them to remove doublespace from dos 6.20
<rqou>
i think he helped them find a random competitor's patent
<balrog>
Lord_Nightmare: they lapsed for nonpayment of fees
<rqou>
idk i have to ask the details
<balrog>
> LZS compression and decompression uses an LZ77 type algorithm. It uses the last 2 KB of uncompressed data as a sliding-window dictionary.
<Lord_Nightmare>
yes but there was prior art to LZS, its just lz77 with some particular window parameters and someone else had used it before stac patented it
<balrog>
that ... doesn't sound ideal
<balrog>
:P
<Lord_Nightmare>
so if microsoft had taken them to court stac would have lost, eventually
<Lord_Nightmare>
i think microsoft decided not to waste like 3 years and massive $ in lawyer fees by just removing it
<balrog>
they replaced it with DriveSpace
<Lord_Nightmare>
yes
<rqou>
in terms of weird computer things my father used to work for everex systems
<Lord_Nightmare>
now that's a name i HAVE heard of
<Lord_Nightmare>
a long while ago
<balrog>
the first external SCSI hard drive I had was from them
<balrog>
centronics 50-pin
<balrog>
I still have the enclosure, the Rodime drive inside died long ago
<balrog>
(we used it to death)
<rqou>
my father worked on their tape systems
<rqou>
and the STEP computers
<balrog>
not familiar with those
<Lord_Nightmare>
did they do floppy tapes? i may have an everex tape drive of some sort
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
mithro has quit [Ping timeout: 260 seconds]
marex-cloud has quit [Ping timeout: 260 seconds]
marex-cloud has joined ##openfpga
mithro has joined ##openfpga
azonenberg_work has quit [Remote host closed the connection]
<azonenberg>
whitequark: you around?
clifford_ has quit [Ping timeout: 265 seconds]
azonenberg has quit [Ping timeout: 258 seconds]
openfpga-bb has quit [Ping timeout: 260 seconds]
digshadow has joined ##openfpga
digshadow1 has quit [Ping timeout: 268 seconds]
Bike has quit [Quit: don't stop]
biot has quit [Quit: Coyote finally caught me]
biot has joined ##openfpga
scrts has quit [Ping timeout: 260 seconds]
scrts has joined ##openfpga
dingbat has quit [Quit: Connection closed for inactivity]
openfpga-bb has joined ##openfpga
azonenberg has joined ##openfpga
Bike has joined ##openfpga
digshadow has quit [Ping timeout: 268 seconds]
mzpx has joined ##openfpga
azonenberg_work has joined ##openfpga
<rvense>
should we register an assembly or workshop for 33c3? it might be fun to meet up in a semi-organized fashion
<whitequark>
azonenberg_work: here
<azonenberg>
whitequark: nvm figured it out
<azonenberg>
had a low-level glibc question
<whitequark>
ah
<jhol>
felix_: did you have time to try the instructions yet?
<felix_>
rvense: the icoboard people already registered an openfpga assembly at the 33c3; i just added myself to that assembly
<felix_>
jhol: no, didn't have time to try that today :/
<felix_>
but it's on my todo-list for the next week
<jhol>
no worries - just letting you know I'm lurking around if yo u need help