
Portals should never be outlined spontaneously but should be well planned and thoroughly tested if the coding is done from the ground up by a development team. It is better to have a confirmation that a sizable load test is also performed on the test base, with clearly identified test cases. I have seen many a systems starting to crumple when the load is put to them. Suddenly the developers and the team who worked on the system will start to get frustrated. The system that was performing very successfully when put to unit testing with a small number of random data will start to pant puff and grind when put to heavy loads with projected number of realistic data.
Planning for features, and integrating features are two different things. For starting a new portal, just the core features should be enough, but should have a portal development plan, and a routine portal maintenance plan. The planning and scheduling should give breathing space to add more services, to switch layout, to give a periodic face-lift and what not.
Currently for most development companies, portal development is nothing but integration of a portal engine with a customized design, or layout. Utilizing the teeming capablities or skinning capablities offered by most of the portal engines can achieve this.