Some points to criticize about LFS

Hui Zhou zhouhui at
Fri Aug 27 13:39:17 PDT 2004

On Fri, Aug 27, 2004 at 09:06:49PM +0200, Gueven Bay wrote:
>The point is not that I want to deviate but the ability to do it. Just the book 
>doesn't give me this ability. You say that advanced users can choose new
>versions. Where in the book is this choosing process described? If I today want
>to use the newest version of glibc which versions should I choose for binutils and
>so on. You see what I mean? And this -- the process of choosing the right 
>versions for compiling a distribution -- is only one of many points which 
>wasn't described in the book. 

I believe the process of choosing the version of certain package is 
to read the mailinglist or bug reports of that specific package and 
build the package by yourself and evaluate the robustness of the 
resulting system -- which requires an advanced user that is capable of 
that. It is rather difficult, and I don't think there is a set package 
version selections that has been sufficiently proven to be the *right* 
versions yet :-).

However, I do like to see some of the results of this choosing 
process to be reflected in the book. For example, for glibc, how about 
the major feature differences between ver 2.1 and 2.2 and 2.3.
>Every configure has many options. I think the work of the writers should be
>to describe which of them I can use to get a working system. But in the book you
>only describe the options for compiling this special system you want.

To describe all the options is not going to be helpful. I once read a 
apache book and a samba book which offers a passage for each configure 
option; the result is very tedious and I didn't want to read even 
once. And many option is just optional, either select or not will 
result in a *working* system, just may not be the ideal working system 
for you. :=)
>>Most of these are taken from the man pages, yes a little better 
>>descriptions could be added. Want to provide them for us?

I think a repetition of the man page is just bad. A phrase such as 
"Please refer to the man pages supplied in the package" is suffice.
How about add something not aim to be complete, but useful. For 
example, how about a pointer to a good tutorial link or good book for 
each packages?
> it wants to give me the knowledge of how this whole thing 

I don't think it will ever *give* you this knowledge. It is too big to 
be contained in any book.
The knowledge comes during your scratch work. Unlike a distro, which one can 
navigate without knowing what is beneath the skin, LFS shows you the 
bone and muscle and encourage you to understand the functions of the 
bones and muscles by googling yourself.

>But no, it doesn't as I am trying to proof. One thing to come closer
>to this is in my opinion to include the hints in the book.

Put the hint in the published hardcopy as an appendix is not a bad 
idea, or is it already been done?
>Hui Zhou wrote:
>>I second your opinions.
>>Although I think I expressed a similar thought on this list a few 
>>months ago. 
>I think the writers just don't see the missing points of LFS. 

Well, even the editors see the points, they may not have that 
interest. We should appreciate what we got, instead of being too 
greedy. ;-)

>And you and I 
>are not native English speaker (at least I am not) so I think we didn't made it

Now I realize we are speaking a little differently.;-)
>J?rg W Mittag wrote:
>>Well, in some cases information is best presented in textual form, in some
>>cases as a mathematical formula and in some cases graphical. E.g. look at
>>Evolutions dependency graph on the LSB website
>><URL:> and then imagine
>>having to grok that in *textual form*.
>Just what I want to say ;-))

>best regards


Hui Zhou

More information about the lfs-dev mailing list