I’m starting to get really confused about these version misconceptions and where they’re coming from.

First of all, about the comment about 0.78 being the second-most downloaded version after the latest. While it’s true, the ratio is huge. The latest version gets thousands of downloads per day in the first few days after release. We’ll usually end the month with around 50,000 downloads of the Windows binary of the latest release, and over 10,000 downloads of the source package. That’s not including all the people who get MAME from elsewhere, clone the repo and compile themselves, etc. MAME 0.78 gets a couple of dozen downloads per day at best – it’s getting less than a thousand downloads per month (split about 2:1 windows binary to source), so it would take over four years to get the same number of downloads as the latest version gets each month. MAME 0.37b5 is usually the third-most downloaded version, with a bit less activity than 0.78 gets. So Haze, I don’t think you need to be concerned than MAME 0.78 installs are growing too quickly. It’s still a very small number of downloads compared to the latest release.

Now these two versions are, shall we say, special because MAME 0.78 is the basis for RetroArch’s MAME 2003, and MAME 0.37b5 is the basis for MAME4ALL. (I don’t know why RetroArch chose to base their “MAME for PotatOS” on 0.78 in particular, I don’t remember it being a particularly significant release. It’s not like they chose to use the last version before the render manager rewrite of 0.106, which is what things like AdvanceMAME and XMAME were stuck on for a while, or the last version of MAME before a C++ compiler became a requirement, like ShmupMAME did for a while). MAME 0.78 and MAME 0.37b5 ROM sets are readily available due to these forks, so I’m guessing people downloading MAME 0.78 and MAME 0.37b5 are taking the advice to “get an older version of MAME that matches your ROMs” that often gets posted on reddit.

What I fail to understand is why there seem to be an increasing number of people getting the impression that the 0.37b5 ROM set is the most up-to-date:
  • It can’t just be from misinterpreting version numbers as decimal fractions. If that was the issue, these people would think the MAME 0.78 ROM set is the latest widespread ROM set.
  • I don’t think it’s coming from RetroArch/RetroPi/Lakka/Batocera, since they don’t push 0.37b5, preferring their 0.78 derivatives. In fact, one for TA’s favourite arguments for why we should be happy with what they do is, “If it weren’t for us, people would be using MAME4ALL, based on MAME 0.37b5 from 2000, not MAME 2003, based on MAME 0.78!” (Yeah, I know – they’re saving us from people using a 22-year-old version by pimping a 19-year-old version. What a huge difference that makes!)
  • I don’t think it’s from the “reference” terminology that’s sometimes used. The top search results for that are 0.78 ROMs (“MAME 2003 Reference Collection”) and 0.174 ROMs (“MAME 2016 Reference Set”). The “MAME 2003 Reference Collection” (0.37b5), “MAME 2010 Reference Set”, “MAME 2015 Reference Set” and “AdvanceMAME 1.4 Reference Collection” (0.106) rank lower in search results. But even ignoring search engine ranking, if they were solely looking at “reference” sets, they’d think the 2016/0.174 ROM set is the latest, because (if you ignore the AdvanceMAME one) those are numbered by year, and 2016 is the latest year there.

So where’s the misconception coming from? What am I missing? There can’t seriously be a resurgence of MAME4ALL derivatives, can there?