An Anonymous Reader writes
"The given deployment details are for my previous employer, who still consults me on a regular basis. I have advocated and deployed Linux there, first as adjunct file SAMBA servers in 1998, for a time on FreeBSD then Red Hat, to entirely displacing NT/Win2k in the server room in 2001, to porting HPUX hosted custom software to Linux as part of my Y2K work, to trialling Linux as desktop, first with Ximian on Red Hat Linux 6.2 to a successful trial with stock Red Hat 8.0 upgraded with the latest Mozilla and OpenOffice.org. Only by providing a proven better solution at the time was anything new adopted.
They currently run one Red Hat Linux 7.2 professional hosting Oracle8 and six Red Hat Linux 8.0 personal running all other services. Each server from a matching retail box, but we only ever bothered to register the 7.2 and one of 8.0 systems, using Red Hat Network web support twice within the first two months of installation and let
the Red Hat Network laps. Red Hat's developers should be commended for the fantastic job they have performed in bringing their Linux distribution to such a high level of stability as a server platform.
Despite not really needing Red Hat Network services, on the server side I consider it a good investment for them to upgrade the Red Hat 7.2/Oracle8 system to Red Hat Enterprise Linux ES Basic or even Standard. However given the existing reliability of the rest of the servers, I cannot yet see any real advantage to upgrading. If the current CIO feels its necessary to have support and upgrade service available for those servers, they may look elsewhere.
Since November 2002, they also run around 42 copies of the free downloaded Red Hat 8.0 as desktop systems, around 20% of the desktops otherwise running mostly Win98se, using Mozilla/OpenOffice for most of those who do not deal with incoming Microsoft Office documents on a regular basis. Netscape/Mozilla has been out default Email/Browser since 1998 and OpenOffice is now deployed on all the Win98 systems. We trialled StarOffice 6.0 and will possibly be looking to the latest release to replace the rapidly aging MS Office98 across the entire enterprise. While all of the PC are currently matched with a legitimate copy of Win98se, almost all of the systems are deployed using Symantec's Ghost imaging, the switch from Win98 to RH8 and back is a mouse click away. After about three weeks of tweaking, the locked down RH8 desktops have caused us much less hassles than the few Win2000pro systems we keep for compatibility. Since the year 2000, the organization have had a purchasing policy which requires all
hardware/printer/scanners etc have compatible Linux drivers, on the possible expectation of a future switch.
It was only due to the free licensing, that we where able to persuade management to trial Linux as a desktops. All the current desktops already have a legitimate license for win98se and Office98. All PC are older Intel P2s to newer Durons, all outfitted with only 128k of memory (RH8 runs a little better than RH9 in 128k). The IT budget has not really recovered from Y2K, which also has meant that MS Licensing6 has not been already adopted either. Despite this, they will have to make a decision in the 2004/2005 period over what form the majority of desktop systems will upgraded to. The leading choices are either major expenditure on hardware and Microsoft XP OR Sun/Suse Java/Linux Desktop on the existing hardware OR a roll your own semi supported solution similar to the current RH8 desktops. A per seat pricing scheme is acceptable but price and price stability over a five year or longer period remains a critical factor.
While the final desktop decision has remained in flux, many of the client side scripting and customized software used before Y2K were dropped and not replace, sticking with the software vendors stock configurations. The result is that very little client side custom software or Office Suite automation is used today as it was when Microsoft Word6 and Office95 were deployed. While it is possible to build Linux application on the Linux Standard Base for future compatibility, we need assurances from the Linux distribution vendor that future Linux releases will still provide binary compatible libraries to standards such as todays LSB v1.3, before such client side work is commissioned or before any major third party closed source software purchases are made. That Red Hat has dropped support for Red Hat 8 and is going to drop support for Red Hat 9 so soon, does not do much to inspire confidence in that possibility. Hopefully Red Hat
will introduce newer product offerings targeting the desktop, perhaps with less support, priced more competitively than the current Red Hat Workstation.
One advantage over Microsoft that Linux/Unix does have in this area is remote X applications. We can keep our custom graphical software hosted on a Linux server behind an internal firewall/proxy, available to any desktop OS that can run an Xserver and it can run until the last server dies after the hardware architecture is extinct. Another advantage over Microsoft is that Microsoft radically changes it's underlying Office suite APIs and scripting system with almost every release of it's Office suite. Hopefully StarOffice/OpenOffice and Java will provide a more consistent platform. That open source licensed products grant effective control to the software to be assured that, in the worst case, the enterprise can go their own way or get together with others to seek support from other parties.
The latter option is not one I look forward to offering as advice to my former employer or my future clients. The high end server and workstation market
alone does not offer enough potential for profitable living for myself as a consultant and contractor or, in my opinion, for Red Hat as a company. Do not forsake the burgeoning Linux enterprise desktop market before it has had chance to truly expand."