Time and inclinations.  If you are mainly guitarist or singer wanting to get your idea(s) down and into your computer, clean and full is good and quick.  And a good preamp can help with that, tho any decent interface preamp can work.  Then add effects to taste, like salt and pepper.
 
Some people prepare the meal before lighting the fire, pepper it early, sauté the garlic and onion before adding the rest of the ingredients, etc.  I grew up in the 60s, listening to music from that era and earlier.  Dallas and KLIF 1190 (AM) pioneered the top 40 format, and most of the music on that station, different as the music forms were, were all well recorded and use the stuff I like today to prefab the instruments.  When I started using hardware during tracking I could listen and go, "Oh, yea, that 1176 on a bass sounds exactly like how I use to hear the bass.  "Or the LA2a on vocals - "Yes, that is the sound I remember and want."  Recording the sound you want can be as exciting as playing the sound that you want, tho it does have pitfalls too.  I've ruined a couple of good takes.  I've also ruined them forgetting to punch record and other stupid, human mistakes.
 
But most of us can keep from ruining a take just using common sense - don't high-pass the bass too high, don't squish a vocal too much and use a mic screen to keep the talent (hopefully) from swallowing your good vocal condenser like a dead 57 during a live show.  A driven transformer tone, EQ and compression are the same - just follow common-sense rules and trust your ears.  If you are an engineer, you want to join in the fun too, just not punch record and keep track sheets updated.  It takes some learning (and understanding the effect of a soft 1176 can have on the bass teaches one what the analog can do going in) but the payoff is nice, capturing a sound you know should work itself easily into the final mix.
 
I don't play any instrument well enough to justify buying good ones (kinda like a color TV for your dog), so I've put money in the front end and record others here at home.  My kick on route 66.  But as far as the sound ...  We were out at the lake and happened to be with a friend of a friend who was a radio guy and music writer.  I had an old song on a mix CD between the Rolling Stones and Crystal Method that came up while we were driving and when asked said it was me.  He thought the vocal sounded great.  This was recorded with an old Presonus Firepod, no fancy pres or comps or EQs.  Made me wonder if I've just wasted all that money on hardware.  But I can tell the difference and have fun so it ain't a waste.  But you don't actually need top shelf hardware to make a great song or good recording. 
 
@

Hi schwa,
I use hardware on the way into the converters regularly, but not always and for specific reasons. Here's how I think of it - I have a bunch of pretty nice hardware, and I like to use it. I also have pretty nice converters (most of them out there are pretty good these days) and so the audio difference between using hardware after tracking or on the way in is essentially negligible. I still do it though, and here's why:
 
1 - Speed. I have some clients that I know are going to have specific things done to their tracks, so I may as well just get them done from the outset. A couple examples: There's one singer that I record pretty regularly. He's a folk/Americana guy, deep voice but needs a little brightness. I know that, come hell or high water, I'm going to push his vox tracks through my TLA-100, cut below about 70, shelf at 120 and drop 2db, and do a wide Q 2db bump at 3.2. Every time. Since I know that, why not just do it on the way in and move on? One more example, female BGV for a band I record. She's gonna come in on my Beezneez (the 87 gets a bit brittle on her), probably on the tube pres since the Great Rivers get a bit honky for her, and I'm gonna squish the dynamcs hard with either an 1176 or a Distressor and make a couple specific eq moves (don't remember what they are right now, I'd have to check their session sheet). Again, every time (at least on her BGV tracks). Doing that stuff on the way in just saves me steps.
 
2 - Performance - Some performers get different outcomes when tracking through comps and EQ's. Another example - I have one female singer, BIG voice, super dynamic, lots of that "whiskey and cigarettes" thing. Running her through the 1176 when tracking changes her performance dramatically, particularly on the softer chest voice parts. When she feels like that part of her voice has more power, stability, and presence she uses it more, and that's a really good thing for her. The 1176 really helps out with that, and since I'm gonna use it anyway, it just makes sense to do it on the way in.
 
3 - Sonar's E.I. plugin. Sonars External Insert plugin is good, and relatively solid, but there are some long-standing issues with it that make me, when possible, shy away from it. Some examples, the soloing behavior on buses, the fact that it consumes stereo I/O pairs on mono effects, no automatic latency recalculation on buffer size changes, and (this may just be my system) it causes weird behavior when count-in is enabled. Now, don't get the wrong impression here - I use it all the time and generally like it, but that stuff does change how I interact with it.
 
All that said, this is just the way I do it, other folks are different and get good results too. Hell, it's not all that  uncommon for me to stay in the box, even with all the crap I have. It's kind of like asking me if I like my Taylor or my Martin better...it just depends on what I'm doing, how I want it to sound, and what mood I'm in.
 
Hope that helps,
Dean
 
 
EDIT - +1000 for what AT said. As usual, he beat me to it and said it better :)

Hi. I usually tend to try getting what I am after on the way in, at least when recording vocals (which is what I mostly record with a Mike ()). Otherwise I usually record as "clean" as possible. 
All the best.

I agree with what most have said in that you just want the cleanest raw material you can get to 'tape'. That said,  a really good quality signal path into the box is a huge  part of that process.  I suggest spending all you can afford on a channel strip to achieve that. There are many budget options, but my personal goto favourite is the not quite so budget Avalon 737 - worth every cent. Mine is about 20 years old, but they still sell exactly the  same unit.
 
In my rambling statically insignificant experience it seems that the only people who rubbish a really good quality hardware pre-amp and compressor are those that have never used one:)
 
 

tenfoot
 
In my rambling statically insignificant experience it seems that the only people who rubbish a really good quality hardware pre-amp and compressor are those that have never used one:)
 
 

I have to agree with this 

I never get out.
 

Straight into my interface, leaving a generous overhead margin to avoid all possibility of clipping.
 
Then I do all FX, including limiting and compression, in the box.
 

Thanks for all of the input.  Lots of nuanced comments to consider, but also common considerations that I was already aware of to a degree. 
 
My interface does not have integrated preamps, but I have other devices that do.  I even have a budget dedicated preamp that wasn't a big improvement to the integrated pres I have used.  The sky is virtually the limit is seems when it comes to mic preamp prices, but I am glad I made the modest investment in my SSL quad pre.  I bought it used, so the pain was even lighter.  But now I am able to record a healthy clean signal from any of my mics or instruments.  (This has sometimes been a challenge in the past). 
 
Others have also said that there's nothing wrong with a little EQ or compression on the way in, especially if you're familiar with the source.  I think I'll continue to experiment with this as my resources allow, but it's ok to work inside the box where my toolkit is bigger right now.

There are usually some background noises happening while I record so I always use a hardware gate or expander-gate. And I usually use 2 compressors, one to catch any sudden unexpected hikes in level that might otherwise cause clipping and the other just to add some gentle smoothing.

@Kev999: If hardware gates misfire, the take is lost.
I use gate (well, expander) plugs in post-processing, and do the sections that fail to trigger properly manually.
 
@schwa: I have a Mackie 1202 VLZ4 which I connect directly to the four line-ins of my interface (Tascam US1800).
 
I add in an Alesis Nanoverb should the artist need a confidence reverb.
 
The signal that goes to the interface is taken directly from the insert jacks, so directly after the input gain. No EQ nor reverb lands on the recording. Bone-dry. 
 
This way I can leave it up to the artist to set their own headphone level, EQ and reverb. Just as long as they don't touch the input gain!
 
The VLZ4 preamps are superb. They are quieter than the self-noise of my Neumann TLM103.
 
I also have a bunch of modified Behringer mini mixers (signal tapped after the input gain) for less critical stuff, like guitar amps and drums.