Re: Hmm. Why no binaries? - an open letter to Mike Pall

["Gerry Weaver" <gerryw@...>]

Hi,

Well... If Mike were to provide binaries for all of the currently supported platforms, he would be maintaining around 14 binary packages. He would then probably need to include debug versions of all of those, which would bring the total to around 28 binaries. Then there would be the library build variations etc. etc.. This would get out of hand pretty quickly as you can see.

I'm curious about what the difficulty was in building the LuaJIT package. Did you read the installation instructions on the LuaJIT web site at  http://luajit.org/install.html? He tells you what you need, where to get it, and how to build LuaJIT.

Thanks,

-G

---

From: Stefan Reich [mailto:stefan.reich.maker.of.eye@googlemail.com]
To: Mike Pall [mailto:mikelu-1109@mike.de]
Cc: Lua mailing list [mailto:lua-l@lists.lua.org]
Sent: Fri, 09 Sep 2011 10:21:46 -0500
Subject: Hmm. Why no binaries? - an open letter to Mike Pall

Dear Mike!

I have to ask you this question directly as it has had me very
puzzled. (Sent to the list in CC because I like things to happen
publicly :])

You have a very good product - an excellent product even. From what
I've gathered and from some of the tech talk you wrote about it,
LuaJIT is an extremely professional undertaking.

You seem like a true professional compiler writer. I know this field a
little and I can tell when I see quality compiler engineering.

That said, I really wonder about your distribution stance with regards
to LuaJIT. Why no binaries...?

This just makes it hard to use your product. It hinders adoption. I
had to ask on a mailing list before I could use your software. Is that
the way you want this?

Why not just publish binaries?

For you, making binaries is a trivial exercise, as you know the
product inside out and you're actually producing them all the time
anyway.

For typical users, making binaries is difficult - it can easily too
difficult to bother, so many will just go and use something else. The
docs for building on Windows are also not that great. It takes
guesswork, especially regarding additional libs (e.g., LuaSocket).

Binaries for Windows are usually extremely compatible with regards to
different OS versions. One binary tends to run on all machines.
(Witness Lua for Windows!) So you wouldn't even need to maintain
multiple versions.

So - where's the point in making users do something that is hard for
them - and easy for you? I really don't get this position of yours.

Best regards and - this criticism aside - many thanks for a free
quality product.

Cheers,
Stefan