RE: [Caml-list] Why don't you use batteries?

["David Allsopp" <dra-news@metastack.com>]

Rakotomandimby Mihamina wrote:
> 09/04/2009 09:37 PM, Gaius Hammond:
> > I am after a language that has the rapid-development of Python or Tcl
> > but with type safety; OCaml is right now the best bet, but it is
> > *very* rough around the edges. The way you install ActivePython is you
> > download it and run the installer and a few minutes later you're ready
to go
> > with everything you need. I'm just reading the release notes for
Batteries
> > now and it starts, you will need <a big long list of things>.
> 
> I must insist on the fact this is probably specific to some OS.
> On debian and Ubuntu, adding one line to sources.list and issuing one
> command line does the trick.
> That is easy, espacially for those already using them.

Using a very simple analysis from
http://en.wikipedia.org/wiki/Usage_share_of_desktop_operating_systems,
97.14%[1] of the computers in the world run an OS which does not have a
"Linux"-style package manager (very sad, but true). Therefore a one-click
installer should be a higher priority than supporting package managers *if*
you want wider adoption of the system at all. Many developers will be using
Windows whether they want to or not (or OS X - though I apologise if OS X
does in fact have a package manager; please subtract 4.59% the previous
number if it does) because they'll be in companies whose IT infrastructure
is Windows, even if they have a few *nix boxes in the machine room "for
those weird developers".

I'm not sure that one is allowed to redistribute the Microsoft C compilers
directly without a license, but packaging MSYS or the relevant parts of
Cygwin along with OCaml and Batteries would create an installer somewhere
between 150-300MB which compared to the 16GB of trial software I downloaded
from the Microsoft website the other day is not that bad. You could even
throw an editor in with it. Microsoft Installer is a command line compiler
which reads text files so it can be targeted from make just like anything
else and the script would pretty much only have to be written once and then
someone would occasionally have to use a Windows box just to build it (and
virtualisation means you can pretend that your Linux PC isn't even running
Windows really). It would also not be too much work to have an MSVC version
which simply explained that you must install the Windows SDK to get the C
compiler (but, just like OCaml's own binary win32 files, you'd still get
bytecode for nothing).

Personally, I compile everything from sources with OCaml - but when I
started out with it (as an undergraduate) I used the mingw installer from
the OCaml website and was up and running in a matter of minutes. If I'd had
multi-step instructions to deal with at that point, I'd have probably ended
up sticking with Moscow ML! My point is that the one-click installer at the
start of the process allowed me to get hooked and then later, when the
benefits were very clear, I was happy to go down the slightly more complex
route of doing it properly. So, for example, once our
batteries-single-click-installer-beginners realise that OCaml is cool and
they want to use other libraries and tools as well then they may well
realise that Cygwin isn't that hard to install and so therefore
GODI-on-Windows isn't really that scary and so they can get Batteries and
lots of other goodies using that instead of the megalithic installer.

Also personally, when I come across a piece of software that offers a
live-CD or virtual appliance as a means of demonstration it makes me look
elsewhere as to me it implies that installation and maintenance is so
complicated that it's made the developers go to the trouble of building a
demo PC (albeit virtual!) just to guarantee that it works properly first
time... but that's probably "just me".

There's an old saying: "If you can't beat them, join them". Standing ranting
on a soap-box doesn't tend to achieve much... 

In vain hope of not starting a flame war,


David (who, by the way, is a realist rather than a Microsoft-apologist)




[1] 4 s.f. seems a tad optimistic on an estimate like that - but rounding to
the more sensible 1 s.f. would mean 100% which might be considered as
over-egging my point ;o)