|
|
| Applications |
| Mailing Lists |
| Miscellaneous |
| Operating Systems |
| Programming |
Re: linux-ipsec: Testing the candidate 0.85 release INSTALL file.
Date: Mon Jun 15 1998 - 15:07:38 EDT
> Here are my raw reactions as I by hand went though the INSTALL file:
And some raw reactions back... No response to a given item usually means "yeah, that needs to be fixed somehow".
>0.
> Does not tell us where a suitable place is nor how much space we
>will need.
Almost certainly superfluous by the time they are reading this file. Probably worth mentioning in README, however. The beginnings of it are now in there.
> 1.
> ...You should give a command line for each
> required step from taring in freeswan and linux...
Including the exact file names, meaning that this file has to be rebuilt for each release? (Remembering also that most customers can't even read this file unless they've already put the release somewhere and untarred it.) I agree with the general philosophy of fully detailed instructions, including commands, but starting it that early strikes me as being more grief than it's worth, and of minimal extra value to customers.
There is also a fundamental contradiction in any set of instructions which tries to provide a cookbook for everything but doesn't spend 10 pages explaining how to cope with "make menuconfig". I suggest that this is a sign that we should have less detail, not more, here -- a reference to an existing "how to build a kernel" document is preferable. Done.
> 3....
> Humm, when I did a make I got not much:
> hugh@north.toad.com: make : Sun Jun 14 13:26:38 PDT 1998
> cd klips/utils ; rm -f version.c ; ln -s ../../version.c
That's actually correct for utils, but admittedly there is still something missing here... :-) Okay, I have now scrapped back compatibility in the Makefile, in the interests of making this work without getting into arcane Makefileology which is too hard to get right. "make" now does everything it should (tested and confirmed).
> 5.
> This single line of text assumes that most folks will know what you
> want them to do. Being more explicit is a good idea.
I disagree -- the level of detail here is just right. It is intended as a warning to the novice ("here be dragons") and an alert to the expert ("something here will need attention"). Cookbook instructions are neither necessary nor desirable.
> 6.
> What does a successful or failed "make insert" look like?
I'd prefer to avoid putting lengthy sample output into this document (several reasons, notably because it tends to get outdated, as witness a later example). A failed "make insert" looks like any failed make with properly-written makefiles; a brief note about that at the beginning may be appropriate.
> 7b.
> The text labels from my last (two, three?) email are still different
> from what I see on the screen...
> The doc will still let me configure with out any transforms. Why am
> I bothering writing this feeback?
Because you've got a bad habit of confusing inaction with indifference. :-) Your earlier comments have not been forgotten, even if they weren't acted on in the hectic hour or so before the first draft release was cut.
> It does not say to finish up the
> configure process but just leaves me haning.
Anyone who needs basic instructions on how to configure a kernel needs more than we can give him. See above.
> 11.
> Reboot how? :r)
See previous comment, although I have added a one-line example.
> I am bothered by the "make all" failing and getting caught by the
> "make install", this seems somehow unprofessional...
Agreed. Too much arcaneness while trying to remain compatible. Fixed.
Henry Spencer
henry@spsystems.net
(henry@zoo.toronto.edu)
Received on Mon Jun 15 15:40:04 1998This archive was generated by hypermail 2.1.8 : Wed Aug 23 2006 - 12:59:20 EDT
|
Contact Us Legal Notices Order Services Online Pantek Home Privacy Policy IT news Site Map Pantek Library |