are Mnesia tables immutable?

[Yariv Sadan]

>
> Google for MapReduce.  http://labs.google.com/papers/mapreduce.html
>
> And, BTW, Google hired Peter Norvig.  Take a look at norvig.com.
>

Very interesting... thanks for the link. I was under the impression
that the Google search engine is written in C++. I do find it pretty
strange that Google doesn't advertise a single job for developers with
functional programming background (at least none that I could find).

>
> Google is very different from most distributed systems.  They don't need
> to be particularly reliable, correct, or accurate and long as they don't
> return anything that is particularly *wrong*.
>
> Again, see MapReduce.  With MapReduce, they either rerun the
> computation, or, more likely, just don't worry about the lost results
> since they have so many other ones anyhow.
>

I got the PDF, and I'll read it now. Thanks for pointing it out.

>
> Is that strictly true?  It is for String, but String is immutable.  What
> about StringBuilder or StringBuffer?  Those will almost certainly have
> some overhead.

String is indeed immutable.  AFAIK, StringBuffer expands by 2x when it
reaches its capacity, most probably using realloc().

I'm not sure about Java, but the SGI implementation of the C++
Standard Libaray has the Rope extension library
(http://www.sgi.com/tech/stl/Rope.html), which provides efficient
operations on very large strings by representing them internally as
trees. I was wondering whether Erlang had a similar library.

Regards,
Yariv