Discussion in 'Business & Enterprise Computing' started by elvis, Jul 1, 2008.
Fuck yeah transport analogies.
In my experience, its not really a case of the sales guy taking the piss. He will absolutely sell you what they think will fit your problem. But if all you have is a hammer, everything will look like a nail. He has no fucking idea - technically - about your problem. And 90% of the time, clients don't know what problem they have. When the sales guy whose main trade is charisma, meets stakeholder who is studid - bad stuff happens - who'da thunk it.
Re: Delivery - yeah well thats a problem. But that typically doesnt' happen within the sales teams control.
Alternatively, you have the most complete version of a solution that fits your budget.
lets the sales guy know how much discount is needed to win the deal. In a prior job with some particularly ruthless sales people I've seen negative discounts (that's right, above list prices) run through before.
To me a good sales rep will give us a quote on what we've specced, but also gives us the option/s and the reasons as to why we should go above and beyond. Its then on the business to pick which way they want to go depending on their risk/budget. That to me separates a salesman from a good salesman. Too often projects i've seen go sideways are because its
a) End of budget time and fuck it just sign something (and its poorly specced so thus goes way over budget)
b) the business expects champagne on a beer budget
c) people who aren't involved in the process (Manglement) spec the software requirements and then the actual end users can't do what they need to.
Where i get involved NSanity is the spec BEFORE it even hits your desk for a quote, i flesh out the random idea that manglement spouts out, and put it into what is actually needed and a decent guide as to things that are wanted/needed/nice to have and how it should all fit together. i don't pretend to be a programmer/db guy/whatever, but i do have a pretty darn decent sense of how it should all fit together, talk together and what it should do that we should get a much tighter quote and be more realistic about what we want, rather then just some manglement brain fart that is emitted with zero clue on what is needed other then it must feel a certain way.
Except you're probably wrong. And don't really understand how performance on our solution works.
You aren't buying enough shit. Because you don't generally get a choice to do this if you do give us a spec.
So all the customers fault. ok.
Except this is where you've failed. Why is management spouting this? if its improvement within your existing environment - then well you should be driving the change with business justification, which if you are half as good as you say you are, should be specced appropriately. If its new products, you should be considered a trusted adviser in your org - and same deal.
The core problem here is that business has you pegged as some disagreeable nerd who can't talk to business and can't talk to finance. Thats *your* problem. I'd say the vast majority of "it guys" can't do this. They have a good idea of what needs to be done (well some of them), but they simply can't articulate that idea to the business in such a way that they can understand it, categorize it in terms of risk/benefit and then allocate funding to it.
Soft Skills are mui importante guys. If you can't talk to people outside of IT, if you can't write a proposal with a business case and risk assessment, then submit it to business and get it approved - you will never ever get any further than blaming everyone else for lack of funding for your environment.
Except you don't know what you don't know. Most internal guys only know their internal dumpster fire. Very few of them actually know whats a good idea, only what has worked in their environment.
I don't debate this at all, the problem is nobody wants to spend the time or money to get it right.
again yep, other than I don't see that as an engineering concern, and no amount of engineering ethics will address it, because the stakeholders will simply fire the engineers and hire ones that deliver what is requested, as best they can, with the time and resources they are given. it isn't reasonable to expect the engineers to fire themselves because of the impatience of stakeholders who don't understand the "pick any two" rule.
(I would hope that the engineering and management teams at Boeing work to somewhat different standards than Nasty Software Inc. I don't even really see the recent Boeing issues as an engineering problem, to my understanding the system works, the failure was in proper sales support for the product.)
Users not reading the manual
although fault on both sides, they made THIS particular model not behave like all the others of the same model, and implemented an extra doo-dad to make it fly the same to get around re-certifying (both the plane and pilots) as a new model, except if this doo-dad fails or users don't know how to use it/what to expect from it, you get pilots piloting them into the ground.
me thinks after this, the regulatory bodies (NTSB/CASA etc) will be more strict on how different a plane can be while keeping it's same type certification.
Yup, hence my lols at "ethical engineering". Ethics isn't even close to the reason anyone does anything.
In order for regulation to work, fines have to be massive before anyone gives a fuck. Literally zero fucks are given about anything *but* money. So this is how we have to manipulate the greedy, selfish world into doing the right thing.
So fuck it, let's up the fines to ludicrous levels. A security breech should be enough to cripple a business (or at the very last have C-level execs ejected in permanent shame when found culpable). Why not? Maybe if profits were really on the line, businesses would give a shit.
against that, the result will be the stifling of most innovation because nobody is going to risk putting anything cool to market if the consequences of failure are so great. forget Samsung and their folding phone debacle, we wouldn't have mobile phones at all because we still wouldn't have conclusively determined that 40 years in the future someone might not die from having one of the things attached to their heads 24/7.
I'm all for responsibility, but look at the current furore over social media - if Mark Zuckerberg can be personally held legally accountable for something some fuckwit with a $50 burner phone posts, you won't be able to post anything online that isn't sanitised, preapproved, run through legal, sent in, sent back, queried, lost, found, subjected to public inquiry, lost again, and finally buried in soft peat for three months and recycled as firelighters. nobody would be willing to take any risks whatsoever, so you'll never see another Bill Gates or James Cameron or Steve Jobs or Elon Musk, and that means we consign overselves to an existence of zero-risk bland.
we all accept some risk in the world for expedience. I don't like that there's any risk whatsoever that driving to work today my wife has a very small but measurable chance of being killed. but the elimination of all risk means the elimination of any meaningful progress, so the reality is we accept that risk for the benefits it brings.
to my understanding, in this case it was very much the classic undocumented design feature, as opposed to a fault with the feature itself.
There's plenty of space to innovate, and risk to be taken in things that don't ruin people's lives. No false dichotomy here.
Up the fines on the shit that matters.
AKA. Our Published numbers make Nutanix look accurate, but if you spec actual work based on them, you're gonna be sorely dissapointed.
What happened to the Nutanix evangelists who used to pop up every time the word was posted?
They were let go, when the world realised HCI is next on the 'is now a commodity' list.
No, it's a huge engineering fail, but the training failure is part of the same issue (I'll come back to that).
Read this article - https://spectrum.ieee.org/aerospace...37-max-disaster-looks-to-a-software-developer (paywalled, sorry)
737-MAX has bigger jet engines than 737-800, requiring that they be mounted higher and further forward. This changed the centre of gravity of the aircraft, and resulted in a dynamically unstable airframe - it is possible for pitching up to result in a runaway condition where the aircraft continues pitching up without control input, which will result in a stall. No previous airliner has been dynamically unstable, and it's a ridiculously dangerous design. So they patched over it - the MCAS system detects conditions that could lead to instability and automatically reconfigures a control surface to maintain stability.
The MCAS system relies on an angle of attack sensor, a small vane sticking out the side of the plane which allows you to read the relative angle between the plane's body and the airflow over the vane. These sensors are known to be fallible (amongst other things, they can get stuck) so on larger aircraft (like a 737-MAX) they mount two. The pilot's instruments read the one on the pilot's side, the co-pilot's instruments read the one on the co-pilot's side. If the measurement looks alarming or wrong (i.e. by comparison to the horizon outside) it's a simple matter to cross-check the reading against the other instrument.
MCAS only uses one of the sensors, and performs no cross-checks of any kind for validity. Is the AoA implausible compared to a gyro measurement of attitude? Doesn't matter, it uses that data. Is the AoA reading too high for the MCAS system to help? Doesn't matter, it uses that data. This is really, REALLY bad engineering. At minimum it should be connected to both sensors and default to disengaging the system with visual and audible warnings if they disagree by more than a small tolerance.
(EDIT: Missed a point in the first pass....) Finally, the original implementation hasd a fatal flaw. MCAS is only supposed to be able to adjust the control surface position by less than 2 degrees (originally only 0.8 degrees, but that was found to be inadequate in testing), but they neglected to track state. Each time the system activated it could adjust the control surface by up to the maximum design deflection ON TOP OF whatever had been added by previous activations. So while it was originally designed to adjust the control surface by a maximum of 0.8 degrees, it could actually move the control surface up to the 5 degree maximum permitted by the actuator.
Back to the training thing. Pilots weren't initially told about MCAS, although there were notes about the system in the manual. Why weren't they told about the system? Because they didn't want to highlight differences between 737-800 and 737-MAX. Why wouldn't they want you to know about the differences? Because the whole point of MCAS is to make a 737-MAX fly like a 737-800 to avoid having the new plane require new type certification. As soon as you call out changed training requirements, you're admitting that the plane is a new type that requires retraining, and avoiding that is the whole point of this stupid system.
So it boils down, as it always does, to what elvis has been talking about - putting the shareholders first over good design.
That's weirdly single point of failure for such a critical system, wtf.
IIRC there's evidence that the Lion crew were trained and followed the procedure that was advertised after the first Max crash. Problem is that it's physically very difficult to execute the manual trim correction when the aircraft is travelling at high speed...and redesigning that system would've required a recertification.
Yeah, if I recall correctly it becomes basically impossible to adjust trim without motor assistance somewhere around 400 knots airspeed. This is probably a bit of a concern if a malfunctioning bit of software has put you into a dive and your airspeed is now well over 400 knots...
Having competed directly against Nutanix - they win everything by;
1. Lying about performance
2. LET ME SUCK YOUR DICK DURING THE SALE PROCESS*
* but you will still end up buying 25% more nodes per cluster to meet your stated performance requirements before go-live is complete
probably trying to buddy up with HPe to try and hide the fact that the Supermicro kit has... security problems.
Dell is going deeper with PE/Rail/VSAN integration. HPe and VMWare are good too - but Nutanix is being left on the outside. HPe obv. doesn't have a skin in the game in the HCI space, so they will play ball with all the girls in town.
how many others are there that worked OK? we'll never know, nobody became an integral part of the landscape as a result. until they fail, in which case we're retrospectively horrified.
and that's all before someone puts a suicidal pilot in charge of a plane. retrospective evidence in both cases, not enough for someone to take action at the time full Monday morning quarterbacking aside. where does that one end? I had a prick of a trip to work today and I'm not in a good mood, do I get pulled off the systems for the day in case I might be (more) homicidal (than normal)? if the guard also had a bad morning, quis custodiet ipsos custodes?
As I understand it, until the MCAS that compensated for the different position of the engines, it was never such a critical system. This has been discussed at length on other fora that I frequent, air-worthiness is one of the far more regulated areas around.