If you were mastering an indie pop rock album and the first song sounds really perfect, but is a little louder than all the other songs except maybe one or two, would you:
 
a) raise the rest of the songs to match
b) say too bad for the first song and take it down
 
We're just finishing up recording today, so it's early in the mastering process maybe, but most of the songs are done and fully mixed, so I can tell that that song is just cranked in comparison...

As you know cranking volume up can cause fidelity problems and dynamics can be too squashed so I would lower the volume of that one track, unless the volumes can be still safely raised. But I'm not a pro so perhaps somebody more knowledgeable in the mastering art will soon comment and aid you better.
kakku

i always vote for less, than more
 
but, it really is a judgement call, and an indication of your ability to master well.
 
no one will ever hear the comparison, except you (if you do your job right, and never let the artist hear multiple versions)

Thanks for the advice kakku.  That's what I was thinking, I guess, but you know when a song is just a little hot and sounds better for it?  Ah well, it's probably just gotta come down...

It depends on how loud it is. If it sounds good (or the way you want) and the meters say it's all good to go, so why not lift the other songs to the same level, assuming their arrangements and other factors allow it.
Do remember, that if you're going to convert the exported project to MP3, you can not master to 0 - -0.2 dB without causing distortion in the MP3. A level of -0.3 dB is considered safe for MP3 conversion qualitywise, AFAIK.

I'd pull the louder one down rather than try to bump the others up.
 
The end user has a volume control if they want to turn it up.

Did you run a global RMS level analyzer? 
An important part of producing a whole album and making sure it is balanced from song to song is to analyze each song and see what the average RMS level is. You can use your ears and stare at a meter, but that is guess work. Knowing a song peaks at -03 Db tells you very little. That could be one little spike.
 
This method is pretty quick, accurate and easy. It also let you know where you are at in the loudness wars. 
A nice friendly safe level I've used over time is around -13 / -14 Db. The war zone songs go as low as -8 Db which will have very little dynamics left. Fine for some music,, not all.   
I use the Wave Lab Global Analyzer tool, but I think there are other wave editors and even free plug ins that will work in Sonar. But seek out a RMS average analyzer. 
 . There is also another measurement standard now being used by pros that is even more accurate then RMS. But for most of us average RMS will work fine. Remember the little VU meter?  
 
http://www.soundonsound.com/sos/feb14/articles/loudness-war.htm
 
 
An important quote from this article: 
 
To work in a loudness-normalised environment, I would urge the use of a BS-1770 compatible meter plug-in inserted in the stereo mix channel, and set up with an appropriate target level. In the absence of a recognised music industry standard at the moment, I suggest that the iTunes Sound Check level of around -16LUFS would be appropriate. The absolute target level is not particularly critical — it's matched loudness with your reference material that's the aim here, along with a reasonable headroom margin to accommodate wanted transients and dynamics.
 
 

Those are some really excellent points, good article - long article.  I bet I've got a decent meter somewhere...I usually just use my ears for relative volume checks - not worrying about the peaks as much as the overall feel.  *But* I've released a few tracks that I wish I would've metered properly.  My most recent albums have been okay, but some earlier albums almost always had a first track that's been too loud - that's probably why I'm concerned about that first track's volume on this current project.
 
The first track always seems to be the hardest to be certain of...all on it's own out there.

Don't decide anything until you can assemble the songs in the rough order you think they'll end up in and listen to the whole thing.

An old hi-fi sales trick is to rely on psycho-acoustics, which mean we tend to regard something 3db louder than something else as "better" (at least, until it gets painful). So they play you the system they most want to sell you slightly louder than the one they're less keen on selling you.

Put the songs in order, then do a rough volume levelling (RMS, with a looser eye on peaks), then you might find that louder song may no longer be the one that really stands out.