2020-11-18 - code dive 2020 notes

…and the code::dive is now officially over. this year was different – due to covid-19 it was fully remote, with presentations being streamed over a youtube (stage 1 and stage 2). additionally there was a slack channel for discussions. since the recording were pre-made, speakers were constantly available on Slack to discuss topics from the talk and do Q&A as the talk progressed, which worked out great. :)

what i am personally missing is the afterparty! :) due to obvious reasons it could not have taken place “the usual way”. i think a WebRTC conference with cameras and beers would be a next great thing here, though! :D anyway – let's hope the next year we will just be able to see each other at the venue and on the afterparty, in person. :)

just some (more or less) random thoughts and notes, taken down during the talks.

Felix Petriconi gave a nice talk on concurrent code testing. it triggered a very nice discussion on the slack, too.

during Maciek's Norberciak talk on embedded systems (great intro, btw!) during slack discussion EE Times embedded 2019 market study was linked. it's nearly 100 pages of graphs and plots i highly recommend you to have a look into, if you're interested in the embedded field (btw: talk also addressed a common misconceptions on what does “embedded system” mean – thank you Maciek for that! :)). some notes i find interesting in the report are:

• 79% of embedded projects are C (56%) or C++ (23%).
• C++ moved from 22% in 2017 to 23% in 2019.
• next year, more new projects are expected to be done in C++.
• around 2/3 of the microprocessor selection criteria weight is not the chip itself or a vendor, but its ecosystem (toolchain, support, etc.).
• FPGAs' vendors were rated very low when it comes to “best ecosystem” ranking (i assume it was normalized to usage group). i hope it will change as more open source FPGA environments show up.
• in Americas and EMEA system-level tools are selected mostly by engineers; in APAC mostly by managers. cultural thing, apparently.

i found a mention on LoRa – a Long Range (up to 10km in rullar area) 300kbps, wireless communication protocol, i completely forgot about. while i do not really need it atm, i think it's a great finding to keep in mind, since ready modules on pinheaders are available for this, starting at 7 EUR price range, which is just great! :D

similarly Bluetooth mesh networking is (finally) there and can be used. modules available. might be a nice thing for a p2p sensory network project. noted down.

another Q&A session of effective code review session by Karol Przybylski resulted in a link to a podcast about code review. according to microsoft research only (?) about 15% of the remarks are about code defects/bugs and about 50% are on maintenance. rest is mostly learning experience – which is great yet often underestimated benefit of a code review process – spreading knowledge across the team. i also linked Sean Carroll's Mindscape podcast episode on mistakes, justification and cognitive dissonance – after listening to that one you'll be surprised how much of what we do is driven by our brain's need to maintain consistent model of the world.

this year Kamil Witecki shared his experiences on need for planned software compatibility. again it triggered interesting Q&A parallel to the talk, on slack. generally you want to plan for both backward and forward compatibility. it never works fine by accident, neither you shall be saved by a framework. it always require lots of though and engineering to make it fly. CyberSec is another interesting dimention here as it often conflicts with backward compatibility – in this world dropping a legacy crypto algo (thus potentially braking some client's) can actually be THE feature.

there was also a talk on feature toggling by Filip Olszewski and Piotr Pietron. FT is a great feature, that is somehow rarely seen outside of banking industry (see: COB-dating). an introduction to feature toggles (aka: feature flags) on Martin Fowler's page, too. last but not least – make sure you're not doing manual deployments, as it can be a real knightmare and companies can get bankrupt.