Alright, the
poll results are in. Thanks everyone for participating!
1) Development for new cams: 57%
2) New features for old cams: 36%
3) Custom (private) projects: 4%
4) Commercial support: 3%
Total: 288 votes.
It's worth noting there was a limit of 4 choices x 25 characters, so I had a bit of trouble finding the best wording. Anyway, here's my opinion on these results:
1) new cameras: this includes primarily DIGIC 6/7/8 cameras (the ones running EOS firmware), but also DIGIC 4+ ones. There are about 20 new models in this category.
Actually, this is what I've been focusing on, since mid-2018, at the expense of DIGIC 4/5 development (
which wasn't funny, btw). Current status: the proof of concept code ("Hello World" on top of Canon firmware) is working on most of these new models (
200D,
5D4,
80D,
77D,
7D2,
5DS/R,
M50,
EOS R, and likely all others), which means by the time ML will be ported to one of these new cameras, 90% of the work will be already done. In other words, the whole group of bodies will be done by "one" common miracle.
And yes - our
latest April 1st stunt is the clearest proof that porting ML on most (if not all) DIGIC 6/7/8 EOS cameras
is actually doable.
Of course, the question is what new camera(s) should I focus on - besides M50 - and most of the suggestions are heading towards 5D4. I prefer to leave this decision for later.
2) old cameras: this one is about maintaining the existing cameras (about 16 models), ideally backporting some of the recent developments (such as crop_rec or Lua improvements).
Based on this vote, I wouldn't declare a code freeze on these old models; I'd just reduce their priority, and maybe allow more flexibility in the codebase, regarding camera capabilities. I'm not targeting every single feature on every single camera. I'd rather remove very old features that are cluttering the codebase, or maybe reimplement them from scratch in a more portable way.
At the very least, the DIGIC 7/8 ports are going to uncover a large number of null pointer bugs (forcing us to fix them), and I want these old models to at least benefit from these fixes; this should be doable without much additional effort. Automated testing in QEMU would be the only sensible way to address this, in my opinion.
Very old models (such as 40D or 5D) are likely in the same category, though I wouldn't expect a high percentage of users. I wouldn't be surprised if one of these gets ML before 5D4, but - should this happen - it will be because their owners were more active on the coding side. Keep this in mind when asking "What about camera X?" as soon as somebody posts an update about camera Y or Z
And IDA_ML already explained
why we shouldn't drop the older models, better than anyone else. Thumbs up from me!
3) custom projects: these would have been custom developments made on a freelancing basis, for paying clients.
I had some similar requests in the past, the most interesting one coming with a $5000 reward. The feature was (and still is) way above my knowledge levels, although it was pretty simple at first sight - but required programming a secondary processor. So... I've looked into it for a while, and then gave up. Maybe it's better to finance research and development time, rather than final outcome. Just saying
If you are wondering: none of these requests resulted in any kind of agreements or payments.
4) commercial support: for the pro users nikfreak was talking about.
It is my understanding the pro users will need somebody available for troubleshooting, if something goes wrong during a paid job (that is, a support contract - idea from
here). At least, I've got 2 or 3 such requests in the past
It is also my understanding the pros need something that just works, rather than a powerful toy. It means lots of time for testing, and I know from past experience that relying on user feedback is not sufficient. It's one thing to publish experimental builds every now and then, and hope the users will report important issues, and a totally different thing (read: amount of time spent on the project) to deliver something reliable.
Automatic testing (in QEMU) could be extremely helpful here, but we are looking at months of full-time work, or maybe years, until it will be functional on DIGIC 6/7/8. It's already working relatively well on DIGIC 4/5, but without an important component - LiveView. That one will also require at least a couple of months of work.
I'm confident I can meet these requirements if there is demand, but they are clearly incompatible with me having a full-time job, unrelated to ML. A one-time boost of $5000 or even $10000 is not going to substitute that, sorry.
In any case, this was the least popular option.
Now, a comment I'd like to address:
My 2 cents: isn't this one of the main things what the ML Bitcoin was set up for, so that development could continue even when lead developers needed pay? I say use that money to pay lead developers to continue the development and when that money runs out do another collection.
That's exactly what we are trying to do, with one difference. Once we use the money (to pay a developer + buy a new camera, or whatever), this money will no longer be available to the project - it will be money already spent. So, I don't think it's wise to spend all the money we've got, on a single developer, without some kind of plan to get it back.
And with the current situation (BTC only), the chances of getting this money back in the next 5 years are - in my opinion - next to none,
with or without 5D4. Why? Look at the
numbers:
- 2013: 2 donations, $22 amplified to $1424
- 2014: 13 donations, $111 amplified to $1925
- 2015: 6 donations, $274 amplified to $9358
- 2016: 1 donation, $25 amplified to $345
- 2017: 3 donations, $46 amplified to $224 (2 after the
3.5K PoC on 5D Mark III !)
- 2018: 3 donations, $47 downgraded to $37 (when most of my time went into DIGIC 6/7/8, including
5D4!)
- 2019: 7 donations, $370 downgraded to $356 (all of them *after* raising awareness)
- Besides BTC, g3gg0 also received €265 in his paypal, to cover server costs (2017 - 2019).
That is, about 14000 volatile USD, collected in 6 years (from about 1200 actually donated).
Currently, website costs are €35/mo (server) + $96/yr (domain) + $30/yr (DNS) + €6.25/mo (protonmail), so about $700/year.
Camera prices (body only, prices from foto-erhardt.de):
- 5D Mark IV: 2500 € (or 2900 € on amazon.de, 3000 € on darty.fr, 2000 € used but in good condition)
- EOS R: 2000 €
- EOS RP: 1350 €
- 6D2: 1500 €
- 90D: 1300 €
- M6II: 930 €
I'm not not saying we should buy all of these, and they don't have to be new. But it does add up, very quickly.
Which hopefully explains why I'm not rushing to take the money and start coding
- a1ex made clear legal advice is paramount. Potential showstopper.
Working on it.
