From: "Octave Orgeron" <unixcons...@yahoo.com> on Saturday, July 25, 2009 12:03 PM:
> Why not add an option at compile/configure time to reference the native > man pages for the target OS? That way it's not a Solaris specific feature, > but an option anyone can use on any target platform?? I think that would > be the best compromise. This is my personal preference. There are non-(Open)Solaris environments where this would be useful, as well. (Such environments are not directly applicable to this discussion and will not be gotten into in-depth by me.) It also removes the stigma of "Sun wants to be different for no good reason" (a complaint leveraged against various of Sun's tools for quite a while now) and replaces it with a lesser stigma of "They chose the configuration options to disable these features" (even though it's the same end-result), if the options are removed. Also, Garrett has raised some very valid points that I agree with. Especially about users opting for removing documentation... Regardless of how valid the desire to do it is, the fact remains that there are business customers who have hard requirements of no documentation kept on their production servers. Having the documentation within the binaries themselves will make trying to get newer versions of Solaris into said customer environments a challenge, at best. Logic is useless against these customers, and it's not an issue of disk space. In addition, I have been observing increasing interest in OpenSolaris as an embedded platform. Due to economies and currently available embedded development platforms, these systems may feasibly have less storage space to work with than even the LiveCD. While it is certainly true that you can get, for example, a 4 GB MicroSD card for less than USD$10.00 these days, they're not exactly a viable option for many/most embedded "consumers". And removing documentation on a device you won't ever actually use the documentation on is low hanging fruit. If my math is correct, a full installation of SXCE 117 includes over 155 MB of man pages alone. Granted, it's a small percentage of total space taken by the installation, but it's also not a small amount of text to plop onto a 512 MB device (most of the pages are in /usr/share/man). But duplicating this in the binaries seems silly and a waste of space. If the precedent is set that all commands in ON should be self-documenting, this may preclude the possibility of an OpenSolaris distribution aimed at truly embedded devices for even longer than it appears the wait will be at this time. If it's specified that "just" things from AST are to do this, I have less of an issue with the space. However, I then have an issue with consistency. How would a user know when to use --man? They'd likely be inclined not to bother, since 8 of 10 times it won't work and they'd be typing "man somecommand" anyhow. And, in addition to the system man pages drifting out of sync with the --man output, Garrett had a great point in that Solaris requires additional sections in their man pages that are not required by other operating systems (e.g. attributes). How have other Unixes with similar requirements dealt with this? Do they patch their custom sections into the source at buildtime or something? Or has ksh93 not been that heavily integrated into any others yet? Also, this reeks a bit of feature creep. While --man, --nroff, and --html can potentially be very useful, where will it stop? Will there be an RFE in a year from someone who needs --ps? Then another in a year and a half for --pdf? And another for --xml? Will they be added? Why or why not? Does anyone have a magic crystal ball to find out? I foresee these options as a slippery slope. Saying right now that "no one will ever need that" indicates that you've not spent enough time with users and management. ;) Also, my opinion toward "We can't disable --man because it will break the test suite" is: Fix the test suite to accommodate the behavior. I realize that this increases the scope of the test suite, but the option doesn't really make sense (at least on Solaris) and the tests should accommodate the Solaris-friendlier behavior. These are just my thoughts on the matter. Make of them what you will. But if I were voting, I'd be voting against this as it stands. However much the *idea* of --man appeals to me: I don't like the look of precedent; I don't like the "trivial" text segments growing; I don't like user confusion (they get enough of it as it is); I don't like not being able to upgrade customers (see the inflexibility, above, about documentation on public-facing systems); I don't like different content depending on how the man page is accessed (they're certain to diverge over time, plus aforementioned Solaris-specific sections); and, I don't like how Garrett's concerns have been brushed aside. Warmest, --Matt -- Matt Lewandowsky Greenviolet http://greenviolet.net/
