How does a T-1 use partial channel traffic for ds1 service

[grover]

Back a life time ago, before the internet (as it exists today), I used
to set up t-1 and ds3 service for a variety of carriers. We were
converting analog mux traffic, divided up into groups, supergroups,
mastergroups etc. into T1 traffic.

Now I see people are using partial T1's for data and internet traffic.
In my mind the T1 is divided up into 192 bit frames, or super frames of
12 frames etc. This divides up nicely for 24 voice slots at 4khz
bandwidth (64kb).

So my question is... can someone explain to me how datatraffic over
takes this rigid structure?

What if I'm downloading a large file of 3mb, and several other people
on my network want to download a file of similar size. (For simplicity
sake- no PVC or PVN, just point to point T1.)

How does that divide out over the T1?

Anyone out there that can explain it to me?