Burton Group recently published a report to clients (see Note below) on portal multiplicity. They make the case that portals are a mature platform that will dissipate as a distinct market in the not too distant future... I say bring it on!
It was and is vogue for large enterprises to invest in portal technology, whether there was a clear vision for implementation or not. Portal Oriented Architectures (POA - attribution to James McGovern for the *OA naming format!) have still not been clearly defined for large enterprises, who are left to fend for themselves and build their own reference architecture based on false starts. The best attempts from vendors still seem to be very infrastructure and SOA oriented. Important as that may be, the portal metaphor is ultimately a user experience paradigm, and any successful implementation must be lead by solid audience analysis and user-centered design. The bright folks at UIE captured the idea, but note that they didn't even use the word portal when they described it.
Vendors are still pitching portal platforms as the cure-all for the presentation layer, when they are too often a placebo. Portals have their place in the EA, to be sure, but they can be a dangerous cancer when applied to every problem. The fact is that portal technology is an aggregation, personalization, and customization platform - they excel when those are the core UI problems to be addressed. They become bloated Golden Hammers when that is not the core problem.
Note: If you work for a large enterprise (any size, really) and are not yet a Burton Group client - become one. Among other things, Burton is one of the few respectable analyst firms that is giving air time to open source initiatives, and seems very comfortable admitting the fact that they play a key role in the enterprise architecture, even for large enterprises (which many still seem to deny).