<balrog>
it is possible since the registers are all documented (unlike ftdi vinculum)
<balrog>
just a mess because of USB standard and chip complexity
<balrog>
I see they're not using an RTOS
<whitequark>
yes
<whitequark>
I would obviously use Rust on it
<balrog>
> firmware will go out of its way to initialize Thunderbolt alternate mode device during boot and run Option ROMs from it
<balrog>
that's less of a problem than it used to be
<balrog>
(enforced signed firmware and IOMMU)
<balrog>
anything PCI-E has that issue :(
<balrog>
that said, thunderbolt is not widely available enough
<balrog>
lack of AMD Ryzen support bugs me :(
<whitequark>
oh, that's annoying
<awygle>
physically I have no confidence in thunderbolt connectors, they feel very fragile
<balrog>
awygle: 2 or 3?
<florolf>
balrog: i haven't seen any useful documentation for GPIF-II, though
<whitequark>
awygle: USB-C connectors*
<whitequark>
I would obviously not use thunderbolt connectors themselves
<balrog>
whitequark: USB-C connectors are much more robust than micro-USB
<awygle>
2. Isn't 3 usb c?
<whitequark>
florolf: GPIF should be trivial to reverse
<balrog>
awygle: same connector, yes
<florolf>
they seem to be doing some reverse-engineering there, so that's nice
<awygle>
Usb c connectors are great
<whitequark>
balrog: hm really?
<balrog>
yep
<balrog>
micro usb 3.0 is better, but it's an ugly connector that's too wide
<balrog>
awygle: thunderbolt 1/2 connectors are obsolete for that
<balrog>
they're still used for mini displayport
<awygle>
micro 3 has fewer connector problems but more board problems
<balrog>
yeah, they're a pain to route
<rqou>
troll: sfp cages
<awygle>
well that and they need serious mounting pins to not just snap off the board
<whitequark>
how many layers do you need for usbc
<awygle>
4 or 6, depending on what you're running it to and how bad its pinout is
<awygle>
that's total layers, not signal layers
<balrog>
but you can just do USB 3.0 - B, right?
<balrog>
(the large square connector)
<awygle>
so ugly
<awygle>
:p
<awygle>
i have a sketched design for a USB 2.0 design that would use USB C
<awygle>
solely because i personally want to go all-C so badly
<awygle>
i also really want to try to do one of those "the board is the connector" designs for USB C
<balrog>
florolf: it looks like cypress describes GPIF-II but it's quite complex
<balrog>
reminds me of the PSoC stuff
<balrog>
:/
<balrog>
awygle: the "tongue" in the connector might be a problem for that
<balrog>
(and might risk wrecking the socket)
<florolf>
balrog: yes, but in very opaque terms (alpha? beta? lambda?). sure, there's probably a straightforward mapping to the graph representation, but i'm glad somebody else is doing the busywork of figuring that out :)
<florolf>
haven't looked at the fx3 in a while, though. the one project where i used usb3 stuff ended up using an ft60x (which, by the way, has firmware and a FT32 core..)
<florolf>
(you need to sign an NDA to even get the instruction format for that processor)
<balrog>
ft60x also requires proprietary firmware? I thought so
<florolf>
balrog: it doesn't require it, it comes prepackaged with some
<balrog>
NDA for ft32?
<florolf>
it's definitely not meant to be user-programmable
<balrog>
sounds like vnc2
<balrog>
hopefully it doesn't suck as much
<balrog>
> you need to sign an NDA to even get the instruction format for that processor
<rqou>
anyways, so that seems to mean that my not-entirely-serious proposal of running 100GBASE-CR2 over passive TB3 cables should be theoretically possible
<rqou>
alright, i really should just test some of this shit
<rqou>
azonenberg: do you have a board-proven footprint for a SFP cage?
pie_ has joined ##openfpga
<awygle>
lol whitequark, you hit a nerve with that SCL question
<whitequark>
what the fuck
<whitequark>
"thunderbolt 3 cables"
<rqou>
so yeah, how are passive TB3 cables different from passive usb cables?
<whitequark>
that's insane
<rqou>
why the fuck is this so complicated
<rqou>
why do my SFP direct attach cables "just work" with no problems from 1G to 28G?
<rqou>
and yes, "just work" because DAC cables are usually exempt from optics whitelists
<rqou>
and SFP direct attach doesn't even have a real spec and it "just works"
<rqou>
wtf is USB/TB doing?
<kc8apf>
TB1/2 put in transceivers with per-cable calibration. TB3 likely inherits that insanity
<rqou>
wtf
<rqou>
why don't SFP DAC cables require this?
<rqou>
and yet they still work
<rqou>
and at higher speeds
<kc8apf>
better cable
<rqou>
and you can also go optical as TB was originally envisioned
<balrog>
kc8apf: but TB3 can be passive too
<balrog>
a question I have is how do passive TB3 identify themselves
<rqou>
i guess SFP+ DAC cables are still higher quality than TB3 cables
<kc8apf>
rqou: only if you were willing to sign away all rights to Intel to manufacture lenses
<balrog>
kc8apf: did Corning do that?
<rqou>
oh, i meant for SFP
<kc8apf>
Intel owned the patents on the lenses used for LightPeak
<rqou>
SFP seamlessly (other than whitelists) handles everything from short passive cable runs to multi-km-long single-mode fiber
<kc8apf>
they shot themselves in the foot with that. Lots of companies wanted to use LightPeak but Intel wanted ridiculous licensing terms
<rqou>
unfortunately SFP is kinda huge and not really suitable for laptops
<balrog>
kc8apf: sounds like firewire pretty much?
<rqou>
lol
<balrog>
they only announced last year that they'd make the specs available
<rqou>
i've occasionally jokingly called firewire "thunderbolt 0"
<kc8apf>
rqou: well, sort of. All SFP fiber modules have a lot of electronics in them tuned over i2c
<balrog>
and AMD still doesn't have support for it
<kc8apf>
SFP direct attach definitely supports more than USB-C and TB3 though
<kc8apf>
that all comes down to cable quality
<rqou>
seriously, somebody here really needs to "just" go and RE the thunderbolt physical layer
<rqou>
but SFP DAC cables aren't even that expensive
<kc8apf>
rqou: not me. Good friend helped develop it
<balrog>
rqou: I know Wendell Sander was involved with Thunderbolt design
<rqou>
tell him the lawyers and apple fucked it up
<balrog>
(before he retired from Apple)
<kc8apf>
oh, he knows
<balrog>
I'm not sure he was involved with Firewire, but it's possible
<kc8apf>
SFP DAC are bulky though
<rqou>
yes, definitely
<rqou>
it's really consumer-unfriendly too
<rqou>
e.g. crap can get into the connectors
<balrog>
you can buy thunderbolt transceivers in qty 1
<balrog>
no docs though (LOL)
<rqou>
you can?
<kc8apf>
remember that Apple USB2 keyboards had a boost converter in them because the industrial designers mandated a cable thickness that caused too much voltage drop from the host
<awygle>
that's weird, were they running 5V electronics?
<awygle>
like "oh no the voltage is 4.2" "but our whole system is 2.5" "oh right lol nvm"
digshadow has joined ##openfpga
<whitequark>
kc8apf: oh
<whitequark>
THAT EXPLAINS
<whitequark>
SOMETHING I WANTED TO KNOW FOR YEARS
<kc8apf>
awygle: they needed to provide 5V to downstream ports
<balrog>
is that also why the keyboard extension cables are keyed to only work with keyboards?
<whitequark>
Apple USB2 keyboards come with this weird ass USB cable that has a notch in it so you cant actually plug it into anything other than Apple computers
<whitequark>
or something
<balrog>
whitequark: which is easy to bypass
<kc8apf>
whitequark: yup.
<whitequark>
fuck
<kc8apf>
Macs will violate the current spec to allow providing spec-compliant power to downstream ports
clifford has quit [Ping timeout: 248 seconds]
<whitequark>
erotically violating the usb specification
<whitequark>
(wrong channel?)
<pie_>
#fpganudez
<rqou>
so, a passive 0.5m TB3 cable is ~$25
<rqou>
a 1m SFP28 DAC cable is ~$50
<balrog>
yeah all the thunderbolt docs are under NDA
<rqou>
so DAC cables really do seem to be higher quality
<balrog>
but you can go out and buy chips
<whitequark>
rqou: its two times longer
<whitequark>
and two times more expensive
<whitequark>
seems about right
<rqou>
nobody wants a 0.5m DAC cable lol
<rqou>
can't even reach top-of-rack with that
<rqou>
a 1m 10G SFP DAC cable is ~$10
<rqou>
although i don't know if that will work at 20G
<kc8apf>
wtf is with this guy on twitter bashing on OSS hdl tool devs
<balrog>
that's the benefit of usb3, spec is available
<balrog>
> @philtor: in general the quality of gcc and clang is on a completely different level (much, much higher quality) than a Vivado or Quartus so that you don't even need to submit bugs because they work.
<kc8apf>
yup
<balrog>
yeah, because gcc has been around for a much longer time :D early gcc wasn't that great (and early proprietary compilers were worse)
<kc8apf>
xlc is still terrible
<balrog>
kc8apf: lol that tweet
<balrog>
"do you know how complicated it is to make all of the userspace tools" -- didn't stop GNU
<balrog>
"do you know how complicated it is to make a compiler" -- didn't stop GCC or LLVM
<balrog>
"do you know how complicated it is to make a kernel" -- didn't stop Linus
<balrog>
right now with open FPGA tooling we're about where we were in the early 90s with open compiler/runtime tooling, I'd say
<kc8apf>
fear. irrational fear
<Ultrasauce>
if only there were some foss toolchain to gain tractino
<kc8apf>
Ultrasauce: that's a great idea. Do you know anyone who could build one ? ;)
<Ultrasauce>
no ~:(
<pie_>
i dont see where the guy bashes on oss hdl?
<awygle>
that's weird, that person seems to be "of our tribe" - their twitter is full of references to Yosys and how terrible FlexLM is
<whitequark>
no one likes flexlm
<awygle>
true. but my point is that i'm surprised to see them hating on attempts to develop open source tools. i was expecting "senior engineer at Synopsys" or similar.
<awygle>
someone whose livelihood would be threatened by open source
<rqou>
threatened? nah
<whitequark>
lol
<rqou>
they can still sell their ip cores
<whitequark>
they still have tons of customers locked in
<rqou>
e.g. see "dwc_usb"
<rqou>
which afaict literally everybody uses
<awygle>
hm, your responses are making me feel an emotion but i'm not sure what emotion it is. i think it is generally in the direction of disagreement though.
<rqou>
this is called "pivoting" :P
<awygle>
well, yes. synopsys would be fine. the engineers who work on synthesis there might not be.
<pie_>
guuys, i dont see where theyre hating on oss?
<awygle>
pie_: you need to move laterally in the tweet thread, which you can't because twitter is garbage.
<q3k>
maybe I've just seen that comic way too many times
<q3k>
while googling for 'PCIe TLP'
<Ultrasauce>
wait it could do gesture recognition too
<rqou>
so i seriously do want to make an "absolutely no setup required" browser-based interface
<awygle>
q3k: i've actually seen that comic, i just missed the "xilly<->xerious"
<q3k>
yay!
<q3k>
I'm not aone.
<q3k>
for some second I thought I landed in this weirdest google results bubble
<awygle>
we've all been in a meeting with "no no no, you make it too difficult" guy
<whitequark>
q3k: yeah it is similar
<whitequark>
except in my case its not xillybus its fx2
<awygle>
and we've all been "who's going to make this magic happen" guy lol
<whitequark>
which is slightly less easy to use but only slightly
<q3k>
i've never had lack with 'fx2' and 'easy'
<q3k>
mostly because I get scared away by the 8051
<whitequark>
I don't mind the 8051
<whitequark>
it's just ... really really slow
<whitequark>
takes forever to do anything at all
<q3k>
as an 8051 is ,yes
<q3k>
what, 1MIPS by design?
<rqou>
awygle: so thinking about it a bit more i have an idea of what should appear in the interface
<rqou>
"just" define (i haven't thought about exactly how to define) a javascript api for controlling the LA and then provide a big "enter control script here" textbox
<Ultrasauce>
just run v8 on the fpga
<rqou>
lol
<rqou>
the control script could theoretically include running migen/yosys/etc. too
<whitequark>
q3k: no it's something like 20-30 MIPS at 48 MHz
<q3k>
whitequark: oh, huh
<rqou>
which can all run in the browser via emscripten as i've already demonstrated
<whitequark>
but the instructions are shit
<rqou>
yes, this approach is pretty much punting on the problem too
<whitequark>
you need like four to copy a byte of data from one place to another
<rqou>
but hey, chipwhisperer gets away with it :P
<pie_>
rqou, is going to make a killing selling to edcators
<whitequark>
more if they're not consecutive
<rqou>
wait, the 8051 doesn't even achieve ~1MIPS/MHz?
<whitequark>
lol of course not
<whitequark>
most instructions are like 2-3 cycles
<whitequark>
jumps are longer
<rqou>
wtf
<whitequark>
its not pipelined
<qu1j0t3>
welcome to 1980
<whitequark>
it's an ancient DS80C320
<rqou>
i thought "modern" 8051 clones are pipelined?
<rqou>
still better than the parallax propeller :P (~0.25 MIPS/MHz)
<whitequark>
wtf
<rqou>
~every opcode on the propeller takes 4 cycles
<rqou>
except jumps which are slower
<rqou>
"but it has 8 cores! that'll make up for it!"
<awygle>
rqou: okay, that's more the kind of explanation i was looking for
<rqou>
it's kinda interesting to me how getting data _into_ a browser environment is so difficult btw
<openfpga-github>
Glasgow/master 0f05a62 whitequark: Update to use latest KiCad libraries.
<openfpga-github>
Glasgow/master 7342e6a whitequark: Add laser cut case, with production files.
<rqou>
er, not into, _out of_
<rqou>
basically all mechanisms for getting data out of the browser *) aren't designed for general data or *) are stuck in "infosec" bikesheds
<pie_>
well xulrunner was a thing at one point
<rqou>
yes i know
<rqou>
i've used it
<whitequark>
lol xulrunner
<rqou>
why lol?
<pie_>
something something electron
<whitequark>
i've used it waaaay before webapps even became a thing
<whitequark>
to write a webapp
<rqou>
i knew about it back then but didn't use it
* pie_
puts a hat on whitequark's nick
<rqou>
i used it in honestly one of the worst possible times
<rqou>
when mozilla was already trying to imminently kill it but electron (or "atom shell") was still being written
<rqou>
there was also a detour through the giant piece of shit called node-webkit
<balrog>
which still exists as NW.js
<rqou>
is it still a giant piece of shit that crashed whenever you accidentally stepped across the boundary in the debugger?
<awygle>
iw rote an ff extension in xul a long time ago
<awygle>
but never tried xulrunner
<rqou>
oh i didn't use xul
<pie_>
arent those all the same as electron
<awygle>
not.. really
<pie_>
i thought it just got renamed 10 times
<rqou>
the thing i was writing just created a giant browser element, injected Components into the global scope of the page inside it, and wrote everything else in HTML+JS
<awygle>
why don't pdf page numbers and document page numbers ever match
<rqou>
lol
<pie_>
awygle, because people are too lazy to do that right
<awygle>
why are graphics always drawn in meaningless coordinates that don't map to anything physical
<rqou>
lol again
<awygle>
(i know the answer to that one, it just annoys me)
<awygle>
(irrationally)
<rqou>
btw fun fact: css coordinates are defined in terms of the subtended angle
<rqou>
so according to the spec, hidpi is supposed to "magically" work
<rqou>
(which of course it doesn't)
<rqou>
whitequark: random question: do you know if there are any plans to make "rustc.js"?
<rqou>
can this be achieved with emscripten already?
<pie_>
"i get it from a guy at the market, he works for Stall-Man, calls it 'Open Source', gives this shit for free, can you believe it? you just gotta help make more"
<pie_>
well, time to make a skyrim mod
<genii>
rqou: "What is he doing with all that peroxide and muriatic acid?" etc
<awygle>
does anyone use non-commercial licenses for open source software?
<awygle>
are you even allowed to do that?
<balrog>
whitequark: the world runs on shitty hacks :D
<balrog>
awygle: yeah and then it's not open source
<awygle>
bleh
<pie_>
non-commercial...for open source?
<balrog>
awygle: best bet is to [A]GPL
<pie_>
im confused
<awygle>
balrog: can you point me at any examples?
<balrog>
MAME, formerly
<awygle>
pie_: like CC-BY-NC
<awygle>
(specifically the -NC)
<balrog>
it's not open source if you restrict who can use it
<awygle>
oh yeah, i think snes9x is that way too
<pie_>
oh /that/ kind of non-commercial
<balrog>
yeah, snes9x too
<awygle>
it's "not open source" according to the OSI and FSF
<balrog>
another example: jslint license
<awygle>
but never mind
<balrog>
awygle: and what generally is considered open source as well
<balrog>
awygle: why do you want such a license?
<awygle>
balrog: i don't necessarily want one, but i want to see a relevant case study of one
<awygle>
i'm increasingly aware of negative externalities to the open source / free software movement
<awygle>
movements, really
<qu1j0t3>
awygle: interested, expand?
<awygle>
oh jslint is this stupid "use it for good" thing
<balrog>
yep
* awygle
lumps it into the "i don't think i need to think about licensing" camp of licenses
<qu1j0t3>
LOL, well this is why real licenses were invented...
* qu1j0t3
remembers Beerware, postcardware, ...
* pie_
mumbles about still needing to get sufficiently ruch off something
<qu1j0t3>
and i think i saw that tweet float past already
<awygle>
it seems like a lot of companies make a lot of money off the backs of open source, while simultaneously leveraging it to pay fewer developers
<awygle>
slash pay developers less
<qu1j0t3>
nobody could have predicted :)
<awygle>
well, yes
<awygle>
but considering the generally anti-capitalist sentiment of the open source crowd a lot of people seem surprisingly okay with this
<awygle>
which kind of baffles me
<qu1j0t3>
well i'd better put my cards on the table, i think inflated dev salaries are extremely toxic to software quality
<awygle>
so i'm looking for examples of attempts to fix those issues
<awygle>
qu1j0t3: "interested, expand?"
<qu1j0t3>
hehe
<balrog>
[16:45:17] <awygle>but considering the generally anti-capitalist sentiment of the open source crowd a lot of people seem surprisingly okay with this
<qu1j0t3>
well, this is basically something i encounter almost all the time at work, because i'm paid hourly
<balrog>
s/open source/Free Software
<balrog>
and you'd be accurate there
<balrog>
that's a big reason RMS hates the term "open source" (or says as much)
<awygle>
balrog: mmm not sure about that. for one thing, many developers seem not to make a distinction between the two.
<qu1j0t3>
when developer time is very expensive, almost all the healthy processes of software engineering are damaged, i find
<awygle>
i think it's something like "FS types are all anti-capitalist, open-source types are mostly anti-capitalist"
<qu1j0t3>
you can't get budget for them, only immediate straight line results
<awygle>
yep, read it
<awygle>
but i'm not talking about the RMSes of the world, or even the *shudder* ESRs
<balrog>
that said, I generally see [A]GPL used for two reasons
<awygle>
just to pick an example (sorry steve), i'm talking about the steveklabnik's of the world, and rust is still Apache (the maximum-capitalist open source license)
<balrog>
first is to discourage/prevent commercial exploitation without give-back
<balrog>
second is to enable commercial exploitation by the originators
<balrog>
while preventing others from being able to do it
<awygle>
i would love to see "non-commercial under this license, easily available commercial license (hit a button), all developers get a share of profits" at least attempted, while acknowledging that it's nigh-impossible to actually achieve
<awygle>
in general, i guess i'm looking for more experimentation than i'm seeing
<balrog>
people end up using [A]GPL for that
<balrog>
(as do some companies like Oracle)
<balrog>
(and GhosScript)
<awygle>
yes but that tends to close development. this is where things like CLAs come in and (can) end up being as exploitative as BSD et al, just from a different group
<balrog>
GhostScript**
<awygle>
qu1j0t3: sorry, i dropped that thread. i don't disagree with you, but i do wonder whether "optimum software quality" is the right metric to optimize
<rqou>
awygle: apparently some guy in a squirrel costume just won a seat for ASUC senate :P
<rqou>
i guess people don't call it ASUCK for no reason :P :P
<balrog>
awygle: you'd have a CLA with any license that has restrictions
<balrog>
so like, with a noncommercial license and no CLA, the rights would belong to the copyright holder
<qu1j0t3>
awygle: Oh, from my pov, it's a REAL problem.
<qu1j0t3>
awygle: it's just hard to convey.
<qu1j0t3>
awygle: general contracting is an amazing education in software engineering.
<qu1j0t3>
amazing + hair raising
<awygle>
qu1j0t3: lol i believe it
<awygle>
i don't know. i want everyone to have good software for free and get paid for the work they do. no one seems to be working towards that goal.
<pie_>
" One of the kids, an overachieving Ivy Leaguer whose Google internship demanded an advanced understanding of high-level mathematics, was completely baffled when it came to using a simple rice cooker"
<qu1j0t3>
when i say it's a real problem, i mean costs are increased, and not by a small amount. costs are high, reliability is low
<qu1j0t3>
i inherit projects that have every kind of dysfunction
<pie_>
“They’re importing children to destroy the culture,” --> "Theyre importing children[...]"
<pie_>
(im not this bad but i could be, maybe im not overachieving enough :P)
<qu1j0t3>
awygle: And then you find, well, developer costs mean that you aren't going to be able to correct it, most of the time.
<pie_>
though in a sense, probably the parents fault as well
<qu1j0t3>
awygle: i mean i'm talking about process things, that have a direct impact on costs of maintenance
<qu1j0t3>
"i want everyone to have good software for free" /// yeah i'm already explicitly inside the commercial, proprietary/bespoke sphere...
Lord_Nightmare2 has joined ##openfpga
<qu1j0t3>
this is code you definitely don't want open sourced, it would be a superfund site
<awygle>
qu1j0t3: lol. evocative
Lord_Nightmare has quit [Ping timeout: 248 seconds]
<awygle>
i think eventually everyone will figure out that software is not that hard
Lord_Nightmare2 is now known as Lord_Nightmare
<qu1j0t3>
sometimes I *wish* it was out there as cautionary tales. some projects have taken literally the worst way to solve any problem.
<qu1j0t3>
some devs can do this in multiple languages!
<awygle>
by then i hope to have sufficient resources to work on things i actually think are important
<qu1j0t3>
awygle: "software isnot that hard" if you have a decent amount of education in not making it hard. It's very easy to make it very hard. Heck, i've been on projects where you couldn't add features because the *CSS* was so bad
<awygle>
well, yes, but that's true of everything. i don't think software is _uniquely_ hard
<qu1j0t3>
i agree there.
<qu1j0t3>
but, the education level seems uniquely low.
<qu1j0t3>
no barrier to entry.
<qu1j0t3>
this is probably more obvious in what I do, since i inherit random code from random small businesses.
<qu1j0t3>
and those businesses never know what they've got inside the box.
<awygle>
hmm, interesting POV
<awygle>
well thanks for the discussion everybody
<awygle>
sorry to take the channel on a wild ride lol
Lord_Nightmare2 has joined ##openfpga
Lord_Nightmare has quit [Ping timeout: 248 seconds]
Lord_Nightmare2 is now known as Lord_Nightmare
<pie_>
a wygled ride
<awygle>
:p
* pie_
shows himself out
<awygle>
go write a skyrim mod
<pie_>
i dont even have skyrim
<pie_>
i keep forgetting, i gues rqou and azonenberg are in the bay area?
<pie_>
not htat i keep forgetting but i never really internalized it
<awygle>
azonenberg is in the pacific northwest
<awygle>
as am i
<awygle>
rqou is in the bay area
<pie_>
ah
<pie_>
i had a hunch the berg wasnt in sf
<pie_>
oh wait seattle was it? well not that its pertinent at the moment anyway
<pie_>
careful or youll end up like bill murray in that one movie
<pie_>
zombieland was it?
<rqou>
pie_ get out of HU :P
<pie_>
apparently ermans like formal methos
<pie_>
methods
<pie_>
germans
<rqou>
they also like vhdl so they can't use any of the f/oss hdl formal tools :P
Bike has quit [Ping timeout: 260 seconds]
Bike has joined ##openfpga
<awygle>
qu1j0t3: do you ever just scrap the crappy internal tools you get handed and replace them with off-the-shelf alternatives? are you empowered to do that? is that feasible in your industry?
<qu1j0t3>
it's mostly the software itself that is bad, but maybe that's to be expected since it's just a random slice through
<awygle>
hm. are these software businesses?
<awygle>
like, software is their product?
<qu1j0t3>
yes and no. they're businesses that depend on software. usually not. often they have SaaS aspirations but that usually comes along with some awareness of the real costs of sufficient investment to do that
<awygle>
i got confused when you said "general contracting" maybe...
<qu1j0t3>
but the other thing i've learned is that a lot of businesses depend on software that barely works and they don't even understand that they are now in teh software business. so they can't manage that effectively and don't know the risks they are taking.
<qu1j0t3>
general contracting, easy to understand, who are you going to call if you're an SME with a software problem and for some reason you don't want to, or can't, call the people who did it? there's a lot of hit and run in the business and relationships end for a variety of reasons
<qu1j0t3>
you could also argue that i see a self-selected set of disasters. maybe so. but i've also worked in other scenarios. a lot of these issues cut across the whole business to more or less degree
<awygle>
i think of a "general contractor" as the guy you hire to renovate your house (unless you're azonenberg) so i was kind of picturing some kind of inventory webapp for plumbers
<qu1j0t3>
if you're not a super technical manager / owner and you hire developers, i think the expectations you have are likely a lot higher than what you're going to get.
<qu1j0t3>
general software contractor, then.
<awygle>
i get it now :)
<qu1j0t3>
sorry :)
<qu1j0t3>
anyway this is kind of top of mind since we just inherited another mess, and trying to figure out how much of a mess the client knows it is, and whether they really are able to pay to make it maintainable. I think a lot of the stuff i see is just developers working without proper supervision or management, too. inexperienced devs, maybe. but there's a lot of it out there. I work on two or four of these
<qu1j0t3>
every year. It's hard work, Aegean stables
<qu1j0t3>
Augean*
* awygle
gives points for the reference
<qu1j0t3>
this can also happen within a big, prestige org. I worked at Amazon. Teams would hand off hot-potato projects all the time, it is proverbial.
<qu1j0t3>
this latest one i'm on is because the old team got either bored or disengaged and needed to punt it.
<awygle>
sure, it happens everywhere. you can't stop and spend six months generating no new features ($$) to refactor everything into a maintainable state
<qu1j0t3>
i'm not calling for a "stop the world".
<qu1j0t3>
i've learned to do this incrementally
<awygle>
those weren't actually as connected of thoughts as they came out as
<qu1j0t3>
the question is, who will pay for it?
<qu1j0t3>
and i'm not talking about adding unnecessary polish, i'm often talking about bringing minimal sanity :)
<qu1j0t3>
anyway, the who-will-pay question brings us around full circle. when the workers are very expensive, it changes the economics of superfund cleanups.
<qu1j0t3>
i.e. direct impact on quality
<qu1j0t3>
in a bad feedback loop, because if you DON'T, development takes longer and is more painful and the outcomes are worse, making it more expensive, round and round
<qu1j0t3>
also a lot of this job is client education. it's hard to deliver these messages and sometimes politically difficult.
<qu1j0t3>
anyway you get the picture!
<awygle>
yup :)
<awygle>
it would be nice to do a kind of tour of a bunch of shops that aren't this dysfunctional
<awygle>
somebody must have written that book, i suppose
<qu1j0t3>
yeah exposure to GOOD practices would help a lot
<awygle>
but they were probably trying to sell agile consulting services so the conclusions would be suspect
<qu1j0t3>
but sometimes i think everyone should do a maintenance stint, like 5 years :)
* awygle
is not good at time, "5 years" sounds synonymous with "forever"
<qu1j0t3>
heh
<qu1j0t3>
i guess i've been doing this type of work on and off for 10 years now
<awygle>
i haven't been doing any type of work for ten years
<qu1j0t3>
occasional greenfield stuff
<awygle>
as i am myself relatively greenfield :p
<qu1j0t3>
i've been coding for much longer, but this particular part of my career has spanned about that long.
<awygle>
as much as i give rqou shit for the academic's perspective i'm aware that i'm not tremendously more worldly
<qu1j0t3>
seems set to continue, as i've got no urge to go back to "permanent" type work.
<qu1j0t3>
(even though this stuff can be very frustrating.)
noobineer has quit [Ping timeout: 256 seconds]
X-Scale has quit [Ping timeout: 264 seconds]
X-Scale has joined ##openfpga
<rqou>
hurr durr
<rqou>
USART1 runs at 2x the baud rate of everything else