bgp design

Does anyone know of any good books that specifically deal with BGP design !?
I have a large OSPF network and plan to implement 4 internet access points
at different parts of the network. I want to control traffic from certain
areas to specific internet access points, but have redundancy to other
internet access points in case the primary route is down. I was looking at
implementing either route maps with attributes to prefer a certain path over
the other or break my AS into Confederations. Is this a solid design plan or
is there a better way of doing this ?

Title: Internet Routing Architechtures
Author: Bassam Halabi
ISBN: 1562056522

Man this makes it difficult to figure out who is to blame/ or
  who to try to help . Looks like nobody at Mae West wants to
  talk to Digex right now . Heck this went all the way to Mae East
  to get back to ELI's connection(s) . Twys , JimL

Type escape sequence to abort.
Tracing the route to ns2.baby-dragons.com (192.103.194.6)

  1 sjc3-core3-fa0-0-0.atlas.digex.net (165.117.58.227) 0 msec 0 msec 4 msec
  2 sjc4-core6-pos1-3.atlas.digex.net (165.117.60.2) 0 msec 4 msec 0 msec
  3 sjc4-core5-pos6-0.atlas.digex.net (165.117.59.93) 4 msec 0 msec 0 msec
  4 dca1-core9-a2-0.1.atlas.digex.net (165.117.55.5) 72 msec 68 msec 68 msec
  5 dca1-core9-pos5-0.atlas.digex.net (165.117.51.245) 76 msec 76 msec
    dca1-core10-pos6-0.atlas.digex.net (165.117.56.54) 76 msec
  6 dca1-core10-pos6-0.atlas.digex.net (165.117.56.54) 80 msec 84 msec 80 msec
  7 iad1-core4-pos4-3.atlas.digex.net (165.117.53.34) 84 msec
    iad1-core2-pos1-0-0.atlas.digex.net (165.117.52.193) 76 msec 80 msec
  8 mae-east.priori.net (192.41.177.143) 80 msec
    iad1-core2-pos1-0-0.atlas.digex.net (165.117.52.193) 80 msec
    mae-east.priori.net (192.41.177.143) 80 msec
  9 a1-0-0--2.cr01.wash.eli.net (207.173.113.242) [AS 5650] 164 msec
    mae-east.priori.net (192.41.177.143) 84 msec
    a1-0-0--2.cr01.wash.eli.net (207.173.113.242) [AS 5650] 116 msec
10 a1-0-0--7.cr02.sttl.eli.net (207.173.114.125) [AS 5650] 168 msec 168 msec 168 msec
11 p0-0-0.gw01.sttl.eli.net (216.190.160.253) [AS 5650] 168 msec 164 msec
    a1-0-0--7.cr02.sttl.eli.net (207.173.114.125) [AS 5650] 172 msec
12 gw-CLICK-NETWORK2-DOM.sttl.eli.net (209.210.81.174) [AS 5650] 172 msec 168 msec
    p0-0-0.gw01.sttl.eli.net (216.190.160.253) [AS 5650] 172 msec
13 gw-CLICK-NETWORK2-DOM.sttl.eli.net (209.210.81.174) [AS 5650] 176 msec
    131.191.8.62 [AS 5650] 172 msec
    gw-CLICK-NETWORK2-DOM.sttl.eli.net (209.210.81.174) [AS 5650] 176 msec
14 sunny.tpl.lib.wa.us (192.103.194.13) [AS 5650] 172 msec
    131.191.8.62 [AS 5650] 180 msec 176 msec
15 sunny.tpl.lib.wa.us (192.103.194.13) [AS 5650] 192 msec * 176 msec
16 ns2.baby-dragons.com (192.103.194.6) [AS 5650] 176 msec 180 msec 176 msec

Hmmm, on second thought I have to ammend the statement below
  it not Digex that's lost It has to ELI then . Ah-heck I'll
  learn to read one of these yet . JimL

Do not do the things too complex. The standart solution is route-pams
with prefix-based, as-path based and community-based matches - why don't
use it? Confederation schema have a lot of hidden stones inside.

Alex.

Btw - I know this book as the _cisco's document_ = was it published by
someone else?