Differences between regression using base R and using rms package OLS models can be created using the base R lm function.
They can also be created using the ols function in the rms package.
Unfortunately, I cannot find anywhere online (including in the ols help page) a summary of the differences between ols and lm, and what reasons there might be in practice to prefer one or the other in particular circumstances.
(I assume that there are substantial differences, otherwise why should the authors of the rms package have created the ols function.)
A similar question arises with the competing mixed model functions in the lme4 and nlme packages, amongst (I suspect) many others.
There does not seem to be anywhere on the CRAN site where comparisons of the features of packages with overlapping coverage can be found. Are such comparisons available elsewhere online?
 A: I'm not quite getting the negative undertone to this conversation, but in brief there are five differences.


*

*ols implements penalized least squares (default = no penalty)

*The default contrast used for categorical predictors is reference cell indicator variable coding.  If you don't have categorical predictors (or if you specify options(contrasts=...), regression coefficients and standard errors from ols are identical to lm.

*ols remembers much more about the data than lm so that graphics displaying the fitted model are easier to draw and certain estimates (e.g., inter-quartile-range difference in means) are easier to get

*rms implements a restricted interaction surface that ols can fit

*Being part of the rms package makes other things such as bootstrapping, model validation, comprehensive anova (e.g., tests of nonlinear interaction) easy to do, $\LaTeX$ typesetting of model fits, etc.

A: I think part of the whole issue here is the nature of R vs. a system like SAS or SPSS or what have you. There are many differences - and many favor R. One key difference is that R is free. That means that the people who develop it are not paid to do so. So, while many of them try to be very helpful, they are writing packages because they think it's fun (or, in some cases, because it helps their careers - publish or perish). 
SAS costs a great deal of money.
One thing you get for that money is people who are paid to do things, including things that are boring. So, for example, in the SAS documentation for most statistical PROCs there is an extensive section on comparisons with other PROCs. In addition, the SAS documentation has big sections on statistics (as opposed to the PROCs), the documentation lists many references, has many fully worked and annotated examples etc.
For example, if you type ?lm you get something that, if printed out, might be 5 pages or so (at a guess). The SAS documentation for PROC GLM (the rough equivalent) is well over 100 pages. The  example at the end of ?lm is 12 lines long (admittedly, there are more complex examples elsewhere). The first example in PROC GLM is about 10 pages.
In addition, and compounding the above, is that R documentation almost prides itself on its terseness; while SAS documentation prides itself on verbosity and completeness. 
I like both SAS and R. But they are very different in many ways, including what you can expect of the people who develop the packages; indeed, at SAS the people who develop the PROCs are usually not the people who write the DOCs; and they certainly aren't the people who answer the help lines. OTOH, here, you  asked a question about rms and you got an answer from Frank Harrell; and you  got it promptly, too. 
A: Please check the handout of FE Harrell here at page 155. 
as in Table 6.1

Function                 Purpose                      Related S Functions
ols           Ordinary least squares linear model           lm

Table shows that there is no difference between these two functions. 
