![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
hirezAfter I posted a two-sentence rant on a friend's FB, I had to stop and think.
Thought one was that all of Rupert's actual friends (ie - not some mental ex-goth failing to make it as a SF writer because interesting computer things and uninteresting life things keep happening, that he last spoke to at least twenty years ago) will consider me a drive-by bell end. Which is how things work on FB because there's no actual sense of how people know each other. There's this amorphous blob of probably-human outside your house that keeps flinging pictures of cats through your window and anyone who stops by for a slice of tea and to parse the time around (Lennon) is going to have to get used to ducking regularly or get as good at gas-grenade tennis as recent French protesters.
I want to see English protesters hitting teargas canisters back over police lines with a good square cut.
The other thought was that my job is a permanent orbit around computers being entirely shite at everything they are called upon to do. Or rather that people are shit at working out what they want the computers for.
Devops, right? Automate all the things! Deploy fast! Fail faster! Devs and Ops (and Net types and DB-fondlers if you want any hope of making progress) working side-by-side on piano keyboards for make great benefit of great christ who writes this shit?
Hardware is by and large awful and will break in as entertaining a way as possible at the least appropriate time. For instance about a week before the warranty runs out. Then it will take $supplier two weeks to mend same, by which time oh dear well yer kit's not supported. So you can either have multiple boxes and load-balance the crap out of everything (or round-robin DNS or...) which increases the likelihood of failure because you've more things to go wrong. Or you can rent service on someone else's pile of computers. Which will expose you to exactly the same set of problems only this time you have to wait for someone else to fix it. Or have three of everything at three times the cost.
OSes wear out. You have got a plan for when your distro is no longer supported and yet another OpenSSL exploit arrives, right? What's that? There's no engineering budget for migrating off ShonkOS-7 to ShonkOS-9 because that would mean rewriting the front-end layer in Grollop since all the devs on the Spon framework left when it turned out the project owner was seen coming out of a Styx gig? Oh dear oh dear. It seems to me that projects exist in two states - either being maintained or being decommissioned. Which is it now, Ralph?
Frameworks. Oh god frameworks.
OS packages are a good idea, though. None of that rubbish with unpacking a tarball on a server and calling './configure && make && make install' a valid method of software distribution. Doing that now would be daft when even toy languages come with their own package management rig. I mean, all you have to do to ensure a repeatable experience is distribute a textfile with yr c0de that lists all the dependencies on third-party libraries and their versions and then run a one-liner to ensure that everything's up to date... You know, does that one-liner look suspiciously like a 'make install' with false nose and glasses? Because that what it looks like to me.
All of this is just papering over the cracks...
... Ah, stuff it.
Thought one was that all of Rupert's actual friends (ie - not some mental ex-goth failing to make it as a SF writer because interesting computer things and uninteresting life things keep happening, that he last spoke to at least twenty years ago) will consider me a drive-by bell end. Which is how things work on FB because there's no actual sense of how people know each other. There's this amorphous blob of probably-human outside your house that keeps flinging pictures of cats through your window and anyone who stops by for a slice of tea and to parse the time around (Lennon) is going to have to get used to ducking regularly or get as good at gas-grenade tennis as recent French protesters.
I want to see English protesters hitting teargas canisters back over police lines with a good square cut.
The other thought was that my job is a permanent orbit around computers being entirely shite at everything they are called upon to do. Or rather that people are shit at working out what they want the computers for.
Devops, right? Automate all the things! Deploy fast! Fail faster! Devs and Ops (and Net types and DB-fondlers if you want any hope of making progress) working side-by-side on piano keyboards for make great benefit of great christ who writes this shit?
Hardware is by and large awful and will break in as entertaining a way as possible at the least appropriate time. For instance about a week before the warranty runs out. Then it will take $supplier two weeks to mend same, by which time oh dear well yer kit's not supported. So you can either have multiple boxes and load-balance the crap out of everything (or round-robin DNS or...) which increases the likelihood of failure because you've more things to go wrong. Or you can rent service on someone else's pile of computers. Which will expose you to exactly the same set of problems only this time you have to wait for someone else to fix it. Or have three of everything at three times the cost.
OSes wear out. You have got a plan for when your distro is no longer supported and yet another OpenSSL exploit arrives, right? What's that? There's no engineering budget for migrating off ShonkOS-7 to ShonkOS-9 because that would mean rewriting the front-end layer in Grollop since all the devs on the Spon framework left when it turned out the project owner was seen coming out of a Styx gig? Oh dear oh dear. It seems to me that projects exist in two states - either being maintained or being decommissioned. Which is it now, Ralph?
Frameworks. Oh god frameworks.
OS packages are a good idea, though. None of that rubbish with unpacking a tarball on a server and calling './configure && make && make install' a valid method of software distribution. Doing that now would be daft when even toy languages come with their own package management rig. I mean, all you have to do to ensure a repeatable experience is distribute a textfile with yr c0de that lists all the dependencies on third-party libraries and their versions and then run a one-liner to ensure that everything's up to date... You know, does that one-liner look suspiciously like a 'make install' with false nose and glasses? Because that what it looks like to me.
All of this is just papering over the cracks...
... Ah, stuff it.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2016-06-11 08:27 pm (UTC)no subject
Date: 2016-06-11 09:45 pm (UTC)no subject
Date: 2016-06-12 07:55 am (UTC)Hence, computing will always be crappy, but not quite crappy enough to be entirely useless.
no subject
Date: 2016-06-12 10:33 am (UTC)Mind, our set up is complicated by regulators deciding to add random extra requirements at irregular intervals. So if Region=this, then use this publishing system, and ensure these 16 items are in place, and that they feed through to inexplicable second version of same text in slightly different format in this way. If Sector=that, then ensure these 24 items are done this way, oh and you will need to display data tables that were never part of this article type. Oh, and if region=this and sector=that, some items will clash, but do the best you can to get 16+24 items in place.
H