Date: Thu, 01 Jul 1999 12:05:55 -0700 From: "Jordan K. Hubbard" <jkh@zippy.cdrom.com> To: mjacob@feral.com Cc: Steve Ames <steve@cioe.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: Lizard... Message-ID: <68739.930855955@zippy.cdrom.com> In-Reply-To: Your message of "Thu, 01 Jul 1999 11:40:13 PDT." <Pine.BSF.4.05.9907011136370.2324-100000@semuta.feral.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> through so-called "spiffy" installs is growing exponentially. Keeping a > simple interface rather than trying to play human engineering with no > real human interfaces lab and a 500K$ testing budget might be better. > Just my 2 cents... I'll shut up now... (I mean, why should *I* beef so > much? I'm not writing or maintaining sysinst...) OK, I'll add a few comments to this. I've looked at the screen shots to lizard and have to say that for a certain type of user, it looks pretty approachable. That type of user is specifically the Windows user, someone who'll have certain assumptions about how dialogs look, what kinds of questions they're asked and so on. It's clear from looking at the screenshots that Lizard tries to be extremely Windows-like and that's probably going to make a reasonable number of people happy; if it's what they're used to, there's really no arguing the point. However, we're also not going for the desktop market quite as aggressively head-on as Caldera is and we have an ISP installed base which would also get very persnickety if our installer started shedding usability in the generic cases in favor of a flashy, focused-on-the-desktop installation. What I mean specifically by that is that we still need to be able to install over a serial line with no keyboard or VGA card plugged into the machine, we still need to support scripted installs (even better than we currently support them) and we need to add support for a number of other features like master/slave installations (one master, n clones), adequate OS and component upgrades, high-level repair/recovery, etc. All the sorts of things that network server users seem to expect and a really whiz-bang graphical install can get in the way of if you aren't extremely careful not to code toward that single goal and have a very flexible approach to the UI. Of course, things can also go too much the other way and sysinstall is a good example of an installer which is significantly constrained just by its UI. The hateful libdialog library has a maximum number of 2 buttons on its standard proceed/cancel dialogs, for example, and there's no easy way to add a "Back" to many of the installation dialogs which could really benefit from one. There are also no tab boxes or panners or any other UI elements for putting more than one screen-full's worth of data up in a useful fashion, that is to say which is easily selectable by the user and with an event model that supports arbitrary call-outs. For the last 3 years or so at least, I've also been stuck in the unenviable position of knowing everything that needs to be done to fix the installer and underlying package tools (e.g. rewrite them from scratch) but having a progressively decreasing amount of time to even think about doing it. About a year ago, I even got a paid contractor working on doing some code and he managed to generate some very promising looking stuff, then my time even for interacting with him became so spotty that he finally got annoyed with me and downed tools on the project until he could get some real and effective feedback from us. My bad, as the current generation says, and it's a major item on my TODO list to spend about 2 days pouring through his code and generating a comprehensive set of comments about where to go from there. Unfortunately, this code is also in the very early stages and represents about 34,000 lines of uncommented C++ and TCL code which requires egcs, turbovision 0.7 and (optionally) Qt 1.42 to build so the learning/testing curve is a bit steep. Every person I've handed a copy to so far has never reported back with anything to pass on to the contractor in question.. :) So anyway, that's why (in a nutshell) that sysinstall and pkg_install continue to get patched and ammended even though we all know how limited and hacked-together they are as tools. :) - Jordan To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?68739.930855955>