While I thought OpenOffice.org (OOo) would read/write WordPerfect docs, I find no control for this, or mentions in the online help.

To quote a friend of mine, "Well, read the [link|http://linuxmafia.com/wpfaq/future.html#ALTERNATIVES|FAQ]." *grin*

That is, there's an alpha-stage filter project, run out of a SourceForge project page, with (if memory serves) development shared between the OpenOffice.org Writer people and the AbiWord people. Progress isn't very fast because there doesn't seem to be heavy effort being put into it. They've bestowed on the underlying library a "version 0.5" moniker for the latest release, which version you can use for command-line conversions. By contrast, the "writerfilter" version, i.e., the OpenOffice.org Writer plug-in filter, currently remains at "version 0.1", which sounds a bit scary.

Me, I've had excellent luck in the past using the wp2latex command-line filter to translate .wpd documents into TeX, and from there to whatever I want using LaTeX stuff, but clearly this isn't what a non-technical office worker wants. Sites that need to do a one-time conversion of legacy .wpd documents into .rtf, .sxw, or something else useful will find the available command-line tools to serve nicely (batching is so nice!), but will have a minor dilemma if they're obliged to have ongoing bidirectional conversion.

Some of the proprietary suites do include robust .wpd filters (Anywhere Office, Ability Linux, TextMaker for Linux).

Rick Moen
rick@linuxmafia.com

jbrabeck wrote:

I started out by doing Paradox programming for them back in '98. Their network is Novell, don't know the version as they have a network support company.

NetWare version can make a significant difference, as the networking protocol stack has evolved over time. NetWare 3.x used the original NetWare Core Protocol (NCP) layers over IPX/SPX transport for machines running its native client software (which, itself, came in several variants). NetWare 4.x added NetWare Directory Services (a very important change!) above NCP, along with abstracting NCP completely from IPX/SPX, allowing variants such (relatively rare) deployments of NCP over IP. I vaguely recall that NetWare 5.x and 6.x have switched to IP transport by default, but please don't hold me to that.

Probably the biggest differentiator is whether you login to an NDS context or not. Of late, Novell is calling NDS "eDirectory", which nomenclature I'm still not used to, as it keeps coming across as a dot-com joke in questionable taste. NDS/eDirectory is a sort-of finely-honed X.500 directory service that I believe has been moving closer to a specialised sort of LDAPv3 structure, over the years.

Looks like you can download the proprietary evaluation version of the Linux eDirectory client software for x86 Linux distributions here: [link|http://download.novell.com/|http://download.novell.com/] They say that you must [link|http://www.novell.com/products/edirectory/howtobuy.html|buy] a non-evaluation copy for any production use. There's also file access via [link|http://www.extremetech.com/article2/0,3973,1157502,00.asp|Web browser](!), which is way too damned peculiar for my liking, and strikes me as of dubious utility over the long term.

There is reasonably good open-source NDS client software for Linux -- [link|ftp://platan.vc.cvut.cz/pub/linux/ncpfs|ncpfs] (NCP filesystem)I vaguely recall that it hasn't quite kept step with Novell's post-5.0 improvements (but, then, there are a huge number of non-upgraded 3.x and 4.x networks out there, so that may not be an issue).

The Linux community as a whole pretty much blew off Novell some years ago as too hopelessly stuck in its own shrinking proprietary world -- which is something of a shame, because they have extremely talented networking engineers/designers. I'm as guilty of this as anyone on the Linux side of things, and so was taken by surprise to find Novell becoming, of late, staunch supporters and adopters of Linux as core technology. It hasn't helped that the main provider of proprietary pieces in the past has been Caldera Systems (of lawsuit infamy). This is of course changing, but there's a bit of an information gap -- and an even bigger mindshare gap -- at the moment.

I have a copy of Corel Linux Office 2000, (WP, Quattro and Paradox for Linux).

Please note that you'll have a difficult time finding many copies of that -- and it would not be lawful to install your one copy throughout your organisation. Remaindered copies can be found [link|http://linuxmafia.com/wpfaq/wp9.html#GETWP9|here and there], but they're becoming more rare all the time.

I personally think that porting Win32 applications only far enough to make them run under Winelib was a dreadful idea ab initio, but I'm a native-app bigot, and you should arrive at your own opinions. ;-> With Paradox, that's all that Corel ever offered, so it was their way or the highway.

It states that the minimum hardware requirement is Pentium 166.

I'll bet a 486 would actually be liveable -- but, to get reasonable performance with ported Winelib apps, not to mention popular desktop toys like GNOME/KDE/OpenOffice.org, you'd need a ton of RAM to go with it, which the retired 486es almost certainly do not have. And, of course, anyone buying new RAM for a 486 in 2003 needs his head examined.

Based on the collective recommendation, I'll suggest to the client that he abandon the idea of "unretiring" the old 486s and purchase new hardware (Walmart has their Lindows machine still at $299.00), and I'll be happy to work with him on getting Linux on all the computers.

Suggestion: Borrow one of those 486 boxes. In fact, stuff it full of RAM scavenged from the other 486en in the pile, as much as it will reasonably hold. Take it home, download one or two Linux distributions, and play with them to see what they're capable of. That will put the rest of this discussion in much sharper context.

Linux in most deployments honestly is not at all CPU-intensive -- unless you're doing large amounts of silly rendering or screen-swooping effects, e.g. with KDE. One of the nicest Linux desktop boxes I have is an AMD K6/233 with about 384 kB of RAM, a nice old Number Nine Imagine128 video card, and a pair of fast 10kRPM SCSI hard drives. I usually run it with my favourite window manager, [link|http://www.plig.org/xwinman/wmaker.html|Window Maker]. The system never seems slow. Just never.

But, generally, I concur with other people's comments that it's difficult to beat the price of new commodity x86 boxes (and a retired 486 of unknown hardware configuration may be too damned much trouble). Only, for heaven's sake, do prototyping. For example, many of the extreme low-end boxes such as the Fry's Electronics $299 boxes preloaded with "ThizLinux" out of Hong Kong, come with VIA C3 Ezra "CentaurHauls" Intel-clone CPUs. Stepping 12 of that CPU is fine, but the equally-common stepping 8 produces errors under Linux on some calculation functions on account of hardware-compatibility problems (and of course you have no idea in advance which chip you're getting). The latter problem can be fixed in any Linux distribution through some rather indelicate library hackery, but the dollar savings isn't worth the hassle.

Rick Moen
rick@linuxmafia.com

I run Windowmaker on a Toshiba Tecra 8000 with a P300, and it's fine for xterms etc. - but run a "big app" like Konqueror and it's pokey. (Yes, I know, a lot of KDE components need to start up to run Konqueror). Still, it's usable. It's definitely NOT usable on a P166 - Netscape on Windowmaker is acceptable there.

The standard Tecra 8000 came mostly in one of two configurations: PII/300, 64MB RAM, and 6.4GB hard drive -- or PII/366, 128MB RAM, and a 10GB hard drive. Sounds like yours is the former.

Laptops generally are a special case, and don't compare well directly to desktop machines (which, to be blunt, is what you were doing by implication), because of compromises in the design intended to keep power consumption and heat-generation low. Those entail slow hard drives and a tendency to run the CPU in slow, lazy modes as often as possible. If you combine a slow hard drive with relatively low RAM, then you are spending a lot of time plowing through slow swap, which dramatically worsens the effect that the slow hard drive would have otherwise. That's consistent with what you report about "big" apps. I'll bet that you'd be radically happier if you were able to justify buying a second RAM stick.

So, in particular, what I'm suggesting is that your attributing the pokiness to the PII/300 is simply incorrect. What you're seeing is attributable, instead, to RAM shortage and slow swap.

Rick Moen
rick@linuxmafia.com

I have to keep reminding myself of the type of machine people mean when they say "retired 486": That would typically be a generic 486DX/33, 486SX/33, or 486DX2/66 with ISA bus only, an antique [link|http://linuxmafia.com/~rick/faq/#486|1994-or-earlier BIOS] incapable of supporting hard drives bigger than 504MB, 32MB of RAM, a single, slow IDE hard drive, a really cheesy Trident or C&T video chipset with from 256kB to 1MB video RAM, and either no NIC or some really cruddy NE2000 clone. And a 15" monitor.

My main machine for eight years running was a 486. Thus: AMI Enterprise III with three VLB slots, five EISA ones, and 16 SIMM sockets for fast-page RAM. Adaptec 1742A SCSI host adapter. ATI Mach32 VLB video card with 4 MB VRAM. Two 7200 RPM IBM SCSI hard drives. One 4x Toshiba SCSI CD-ROM drive. One Yamaha 4X CDR burner. Two 3Com 3C509B ethernet cards. It used a NEC 4FG with effectively a 16" display until I retired it and got my 17" ViewSonic. 32 MB RAM, eventually doubled (about halfway through its service life).

That machine was beefy enough to be my SMTP + Apache + ftp + ssh/shell server for all of my remote users, plus at the same time do local X11 desktop duty for me (fvwm2 window manager), plus at the same time burning CD-ROMs. (Mind you, this was in the days before Linux word processors hogging 73 MB of RAM just to load.)

It was more than acceptable for all of those simultaneous functions during all eight years of service, because I designed it for high I/O in Unix OSes, throughout the hardware. When I say "in Unix Oses", I mean that it was a balanced design, not skimping in some areas of hardware and splurging irrationally in others where Unixes don't benefit very much, resulting in an overall Unix system with sub-optimal performance for the dollar.

The reason I mention all this is the sucky-Linux-system-for-the-dollar syndrome's still with us, now being driven by a new gang of idiots: gamers.

That pretty much was inevitable, as a matter of market economics: The hardware market collapsed, and the only growth area remaining has been the sort of system marketed to teenagers of all ages who fancy themselves gamer D00DZ. You've probably seen them in PC clone shops, and wondered what the hell they were: Transparent compact Plexiglas cases with handles for carrying, neon tube lighting, garishly bright coloured motherboards, ridiculously overpowered CPUs (often 2.4 GHz Athlons) with oversized heat-sinks and fans, NVidia 3D video cards requiring proprietary X11 server software -- and sucky, pathetically under-engineered I/O in every area other than video.

The kiddies who favour these things are basically MS-Windows gamer types who wander over into Linux mailing lists and start giving dumbass hardware recommendations: I see that happen all the time, posting recommendations of poor-value hardware that's poorly balanced for Linux use -- way, way top-heavy on CPU power (and consequently on power draw and heat dissipation), while pathetically bad at I/O (especially mass storage).

My systems, including my former 486, get spec'd to give excellent performance for the dollar over many years of projected service life -- and also to get good parts longevity, through low heat build-up. The ones the kiddies are recommending are overpriced for their real-world Linux performance today, will tend to die early (because of heat), and won't wear well over the years to come. And they're making the same type of error that people made earlier, just with fancier and newer hardware.

Rick Moen
rick@linuxmafia.com

jbrabeck wrote:

A county office wants to look into converting their desktops to Linux. The M$ apps that need to run are an Accounting Package (Dynamics?, a GP offering) and Paradox.

Those will probably run on WINE (open source) or one of the [link|http://linuxmafia.com/wpfaq/future.html#ALTERNATIVESWIN32|souped-up proprietary variants] thereof. However, why not leave the accounting Windows box running Windows? I've never understood the "We're a single-OS shop" mindset.

They are currently using Word Perfect Office for WP and spreadsheet.

If they really like it, they can probably run that Win32 version under WINE, too. I've heard favourable reports. Check on the Corel support newsgroups (linked from my WordPerfect for Linux FAQ).

I have found (I actually have a copy) Corel Linux Office 2000 which contains WP, Quattro and Paradox. But Corel no longer is active in the Linux world.

Yes, curiously enough, they stopped producing applications for Linux (but not for proprietary Unixes) at pretty much exactly the same time as an infusion of cash from Microsoft Corporation. WP9 for Linux/Corel Linux Office 2000 [link|http://linuxmafia.com/wpfaq/wp9.html|was never great, but can be improved]. But, honestly, recompiling Win32 source code to run under Winelib (which is what what Corel did with "Linux Office 2000") is a half-assed way to produce applications for Linux, and is a poor alternative to true native applications.

I could rewrite the Paradox app in ??? and have them use Open Office for WP...

OpenOffice.org isn't a bad suite, especially if you need maximal compatibility with Microsoft Corporation's really dodgy document file formats. The word processor AbiWord's .doc compatibility is about comparable, and AbiWord is much lighter / faster than OpenOffice.org Writer, e.g., loading in 6 MB RAM instead of 73 MB. The spreadsheet Gnumeric occupies a parallel relationship to OpenOffice.org Calc.

But it all depends on what you need. Take your time, try things out, and see what works. For heaven's sake, don't attempt a large-scale deployment without doing pilot projects and agreed-upon acceptance criteria. If you run up against roadblocks, don't give up; ask in places like this one.

...but I don't think I can get them to change their accounting software.

So, don't. Leave that machine alone. It's doing its job.

Which version of Linux?

As the saying goes, that's a religious question. ;->

Pick one; you can't go too badly wrong. You can become a Linux-distribution bigot later.

Open Office?

OpenOffice.org is very useful in providing one-stop shopping for word-processing ("Writer"), spreadsheet work ("Calc"), presentation graphics ("Impress"), and vector graphics ("Draw") -- plus a generic database-access and query interface that can talk to pretty much any back-end database engine via ODBC, JDBC, or specialised middleware.[1] Please note that the name of the program is technically "OpenOffice.org" (abbreviated as "OO.o") rather than "OpenOffice", because the latter was already established as the name of a major Linux VAR in the Netherlands. Also, please note that the ostensibly separate OpenOffice.org applications are actually a single binary that does all of those functions.

Database (must support record level locking, multiple user).

A brief history of databases on Linux:

For most of the last decade, pro-SQL bigotry has been the rule on Linux, which if anything accelerated when all of the big proprietary SQL database vendors but one (Microsoft Corporation) suddenly released supported versions of their SQL engines for Linux in late 1998. The best known open-source SQL engines are MySQL and PostgreSQL. Both have gotten to be quite good; MySQL has more widespread technical documentation and information, while PostgreSQL is technically more solid but less widely used. (Reason, in part: It used to have severe performance problems, which is no longer true.)

Because of the fixation on SQL, and corresponding monopoly over mindshare, other categories of database (xBASE, casual databases) have been a bit slower to develop, but over the last five years have really taken off. I have a [link|http://linuxmafia.com/~rick/linux-info/applications-databases|listing] of known options in all three categories. Sorry to not able to give guidance about which to use for what purpose, but at least the names and locations of those projects should be a good starting point.

Most of my experience is with MySQL and PostgreSQL. Just to be really clear: Both of those are engines, which by themselves furnish only command-line query interfaces. It's intended that you'd use them with separate user-end applications such as phpMyAdmin (Web front-end using PHP) or OO.o -- which is (mostly) what I do.

Caveat: You haven't mentioned what the Paradox application(s) do(es). Duplicating the functionality of a developed application can be a very non-trivial exercise. Beware of "iceberg obstacles" in this area: 9/10 invisible until you run into them. You may not even be able to determine what's the best software to use until you've studied the problem carefully.

Perspective: Although OO.o is a useful tool, its disk space and (especially) RAM requirements are regarded as absurdly high by the standards of Linux applications. If you can devote 200-300 MB of disk space just to housing this office suite and 73MB+ of RAM to run it, then great -- but it is considered a hog.

Accounting?

[link|http://linuxmafia.com/~rick/linux-info/applications-accounting|list]

Reason for conversion is cost, they don't want to spend money upgrading HW AND SW. "We have a 'bunch' of old 486(!) computers that we don't want to upgrade."

OK, whoa, there. This raises the question of hardware adequacy -- not the CPU (which is fine) but rather RAM and disk space. And, in what might seem a small-p paradox, the Linux distributions and desktop environments commonly urged onto new Linux users tend to require (by Linux standards, though not Microsoft ones) gobs of hardware beefiness in exactly those areas. If you hand me, an expert Linux user, a 486DX2/66 with a 504 MB hard drive, a good video card housing at least 4 MB video RAM, and 64 MB system RAM, with some care I can craft a reasonably "friendly" Linux desktop box. It will not have GNOME or KDE; it will not have OO.o; it will not start up a bunch of little-needed daemon processes that far too many distributions auto-default to running with no real reason. It will have a painstakingly selected set of lean applications, instead of the traditional cornucopia of thousands of apps. It might feature [link|http://www.plig.org/xwinman/icewm.htm l|ICEwm] instead of the more feature-rich, RAM-guzzling window managers.

You as a relative newcomer to (desktop) Linux probably wouldn't even know how to go about doing that, unfortunately, for lack of acquaintance with the problem space. On the other hand, if you think you might be up to the challenge, go for it, with my congratulations on your crusading spirit -- but, please, leave yourself large amounts of time for testing and prototyping. But be aware that low-end hardware is a constraint that means you can't just grab a random Linux distribution and, with no idea what you're doing, conduct a "forehead install" (proceeding through the prompts by banging your forehead on the space bar as needed), and expect to enjoy the results. To get a satisfactory outcome under those constraints, you need to know the applications on your Linux distribution of choice; you must know how to do a custom installation to pick and choose the ones that are appropriate for your hardware; you must know how to configure the startup configuration to control what processes start up and what graphical environment the user will have.

The apparent paradox is that "free" (and thus low-end) hardware and "free" Linux mix well only if you have significant desktop-Linux expertise. Unfortunately, newcomers are often lead down the garden path of doing Linux on low-end hardware without expertise: They try loading RAM-and-disk-guzzling "friendly" applications onto a 486 with 96 MB RAM and a 2.1 GB hard disk, run out of disk space while running the installer, try again with some subset of the thousands of apps offered, succeed in completing installation but then find the resulting overburdened system to be dog-slow, and give up on desktop Linux entirely.

(For similar reasons, newcomers tend to insist on dual-booting their proprietary-OS boxes with that OS plus Linux, and then act all surprised when multiboot configurations turn out to be complex and sometime problematic. Surprise: That's always true of multiboot.)

About video: X11[2] on less than 1024x768 resolution is just not satisfactory. It just sucks in diverse, individually annoying ways. Therefore, you need 17" or greater montitors, and video cards holding at least 2 MB of video RAM, bare minimum. (At 16 bits per pixel = 2 bytes per pixel, imaging a video frame at 1024x768 requires 1024x768x2 = 1,572,864 bytes of video RAM. Obviously, 24 bpp or "true colour" looks slightly better, and requires 2,359,296 bytes, which is 4 MB when doing the necessary rounding-up.)

Don't get talked down to 15" monitors. Don't get talked down to 1 MB video cards. They "work" for small and long-term-unsatisfactory values of "work".

Or should I convince them to upgrade HW?

As always, it's a cost-benefit proposition. Unfortunately, it's pretty common for some manager to idly suggest Linux because it's "free" and not take seriously the project's requirements for study and prototyping, and assign to the project a tech with no experience in the area, to do it in his spare time with no funding and short deadlines. The moment the "free" alternative turns out to have caveats and a learning curve, the mandate for it disappears -- because there never was any management buy-in, and what was perceived as "free" was in the final analysis valued at cost. Don't fall into the trap of being that tech: Work out a proper scope-of-work document. Define milestones for acceptance in pilot deployments. Get signoff on your achievement of those milestones and deployments of them. Trust me, you really don't want to agree to head a project with poorly defined goals and acceptance criteria, with no real buy-in from management.

And don't even start until you've played with a few Linux distributions and gotten to know desktop systems somewhat deeper than just the button-pushing level, e.g., buy and learn from Matt Welsh's [link|http://www.oreilly.com/catalog/runux3/|Running Linux] tutorial. Since you've done server/firewall work, you already have a major leg up on that.

[1] In the area of data access and (especially) forms, the just-released OpenOffice.org version 1.1 is much, much better than the prior 1.0x versions. Almost any Linux distribution you install today will furnish the latter; therefore, expect to have to pull down upgrade packages.

[2] Also known as the X Window System, and often incorrectly called "XWindows", i.e., the display engine / mechanism to image graphics on most Unixes. Please note that X11 does best with 3-button pointing devices. 2-button devices are more-or-less OK; in that case, you emulate the missing third button by chording the other two.

Rick Moen
rick@linuxmafia.com