Last year Linux celebrated its 20 year anniversary, as Linus began writing it in 1991 and released it that same year. For myself, I'd have to think more deeply about it to determine exactly what year I first started using Linux, but it was definitely between 1993 and 1995. I installed my first Linux on a non-brand-name laptop with the Slackware distribution. It required many, many 3.5 inch disks to install (I forget how many disks it was if you wanted to install X-Windows - it was a lot). It used LILO as the boot loader.
My next Linux desktop I began using in February 1996. It was at a small startup - so small actually that initially the machine doubled not only as my desktop, but as a server of sorts. The machine had a pre-1 kernel, but the previous sysadmin upgraded it to Linux 1.X the first day I began working. The /proc/meminfo file in Linux had recently had a cached column added to it, which broke top and some other things, although the previous sysadmin dealt with that snag as well. FVWM was my window manager. The machine was susceptible to the "ping of death", and this was before firewalls, NAT, iptables and the like were widespread in use.
Linux as a desktop has come a long, long way since then. Even in the past few years, the Linux desktop has come a long way. A lot of people have done work on this, but Ubuntu has been a big part of this, and is what I currently use on my desktop. One example - when installing my new Ubuntu system, it sets up what is necessary to get the disks and network connection in order, and then it simultaneously starts downloading needed packages from the Internet while I go through the rest of the system setup menus. If I finish all the setup before everything is finished downloading, I can cycle through a promo which shows me which features Ubuntu has. This is the kind of thinking that has been needed for Linux on the desktop - every previous Linux install I remember consists in me doing system setup, and then I have to wait for the downloads to start and finish. Even though it is a little more of a pain to implement from the developer's point of view, Ubuntu gets those downloads started as soon a possible, and I don't have to wait that extra time that I'm doing the rest of the system setup - and if I finish before the downloads, I have a promo to look through where I can learn about the system. Not a big thing, but an example of the kind of thinking needed.
To a point others have addressed - nowadays there is server, desktop, laptop, tablet and smartphone. Linux is dominant on smartphone, and has a very healthy-sized chunk of tablets. With the Nook, the Kindle Fire, as well as the pure Android tablets like the Samsung Galaxy Tabs, the Xoom etc., and this year Ice Cream Sandwich tablets coming out, I'm confident Android (and thus Linux) will grab more of the tablet market. What I have not seen mentioned here (maybe I missed it) is that tablets sales have been cannibalizing desktop sales, and the traditional desktop is growing less relevant over time for the average consumer. Just like people here are saying desktop Linux is mostly only relevant for techies like us, in the future desktops may become mostly relevant only for techies like us, or people doing things like 3d modeling and the like. With a Bluetooth keyboard (foldable or solid) and a tablet, you can already do a fraction of what the desktop can do, and that will only increase with more powerful tablets, new software and so forth. And Linux is a big player in this domain. I think the efforts to improve desktop Linux are good, but I think efforts to promote Linux are currently more productive in the areas where it already has a significant position: servers, smartphones and tablets.
Right now Apple and Android are in a real race for dominance over smartphones and tablets, and Windows is throwing its hat in the ring with Windows 8. I think Microsoft realizes it has lost the smartphone race for the next few years, and will be concentrating on tablets, where the race is still open to some extent. I'm sure their tablets will come with Word, Excel and so forth, which will give them an edge in this space. The app which can do some Word/Excel functionality on Android, Documents to Go, costs $29.99. And people need it, it has over half a million purchases. There is really no good, say, open source Excel-compatible spreadsheet for Android right now. I should know, I wrote the most complete open source Excel-compatible spreadsheet for Android that is out there right now. And "most complete" is strictly in a relative sense - it is really a prototype that I put aside once I hit the stumbling block of trying to load Excel 2007/2010 OpenXML files (it can do pre-2007 though). It has had some interest from other developers, but interest has not been that high. Anyhow, in terms of efforts to promote Linux, you want to concentrate on the more productive channels. I think it is good to push forward on all channels, but I think writing software (even non-FLOSS!) for Android tablets is currently the best place to concentrate efforts from a purely promotional standpoint.