Open jgabry opened 2 days ago
the accompanying text clarifies that if you're starting a new project, use CmdStanR. what is wrong with the term "legacy"?
the accompanying text clarifies that if you're starting a new project, use CmdStanR
If you're starting a new project that involves pre-compiled models you still need to use RStan (or you need to distribute the pre-compiled models yourself, which is very uncommon in the R world). So I was just saying (perhaps I worded it poorly) that "legacy" makes sense except for in that scenario where RStan is basically the only option. I wasn't sure where to convey the info to users that RStan is still required in that use case, which isn't an uncommon one.
I guess we mean different things here by "project" - what I mean is if you're planning to use Stan because you want to analyze your data, you should use CmdStanR. the number of people building software is still very small compared to the number of people just using stan.
I think we can say cmdstanr is recommended for new users without saying RStan is legacy - to me, that implies it’s unmaintained, which it isn’t
I think that new users should be strongly encouraged to use CmdStanR, unless
changed the description of RStan as part of PR #19
What if instead of "legacy interface" we say "original interface"?
Oh just saw your PR, I'll comment over there.
I said this in my brms PR, but I closed that one, so saying here: I hope I wasn't coming across as combative, I didn't intend to. The new site is fabulous!
I understand why you refer to RStan as the "Legacy R interface to Stan" since it doesn't support the latest Stan version(s). But using RStan is still the only way to get pre-compiled Stan models on CRAN, so for the ~100 packages that depend on RStan (and all their users) it's not really a legacy interface, it's the only interface they can use. Given that, I'm not sure how we should refer to RStan. Do other people have thoughts about this?