1.4-641
	Try to byte-compile the estimable functions.
	Fixed a bug which prevented compilation with -DNOTHREADS

1.3-572
	Fixed a memory allocation bug in the C-code, use malloc(), not Calloc() in threads.
	Detect number of cores at load (for unix'es that is, don't know how to do on windows).
        Made kaczmarz parallel, much better than the silly foreach-stuff.  Needs cleanup of C-code.
	Tried various methods for getting the right variance for the bootstrap.  Seems to work now
	Implemented a zero-mean estimable function in efactory(obj,opt='zm') and a 'zm2'.
	Added an is.estimable() function to check user-written functions (and my own) for estimability.
	
1.3-512 
	Bootstrap standard errors with Kaczmarz, user defined
	estimable function which defaults to something similar to what
	lm does.

	Give progress report every 5 minutes.

	Allow group interactions of the type G(id*firm).

	Demean y together with x, improves parallelism.

	Handle multi-collinearities in the ordinary covariates.

1.2-452
	Remove duplicate rows in the Kaczmarz method.  

	Different convergence check, both in demeaning and Kaczmarz,
	using convergence rate summed up to infinity.
	
1.2-444
	Implemented the Kaczmarz method for solving the factor-system, made it default.	

1.1-434
	Changed the argument name 'model' of felm to 'formula', also call model.frame()
	in the parent context, so one may get away without a data.frame.

	Updated and expanded documentation.

	Added a summary.felm function, and a corresponding print.

	Added syntax of the type y ~ x + G(id) for the grouping factors.
	The fl argument to felm still works.	

1.1-427
	Changed the print and summary functions. Class "fe.lm" is now "felm".
	In particular the coefficients-entry from felm is now a vector, not a matrix.

1.0-418
	Initial version on cran