MOST FEDERAL aid to the poor is in the form of entitlements. That's a valuable attribute; it means the money doesn't depend on, and need not compete with other programs in, the annual appropriations process. Everyone who qualifies and applies gets the statutory help automatically. Congress can still alter the cost, but to do so it has to alter the underlying rules.

Housing assistance is one of two major exceptions to this rule. (The other is Pell grants to needy college students.) Nationally there are only enough subsidized units to accommodate a fourth to a third of the eligible population, and the growth (or not) of the subsidized inventory each year depends on appropriations. In recent years the eligible population has grown faster than the inventory; the programs have lost ground. Geography is also a factor; more even than aid to families with dependent children and Medicaid, which vary by state, the availability of housing aid varies by locality.

The House version of the housing bill now in conference would take a first step toward reducing this randomness. For one small category of recipients, it would make housing assistance an entitlement -- families with children who, without housing aid, would either have to be placed in foster care or could not be taken out. Rep. Charles Schumer (D) of New York, the proposal's author, says that the category is narrow and the priority it confers would avert great harm at small net cost (whatever would be spent on housing minus whatever would have been spent on the foster care plus the longer-term costs of family disintegration). Why not do it?

The objection, from the administration and others, is that this could be the opening wedge in a costly fundamental change in the housing programs that should not be casually enacted. In saving one set of families in need, the critics say, Mr. Schumer would be skipping over others equally needy. That leads them to the belief that, once the precedent was set, the entitlement would be steadily enlarged; they say this is a ragged way to proceed.

The proposal's defenders don't so much dispute this as say they have to play the legislative cards they're dealt, that Congress often changes programs incrementally and in fact has usefully expanded Medicaid in recent years in just this fashion. They also say the risk in the present context is hardly that the government will end up spending too much on subsidized housing, but rather that it will continue to fail to meet the need. All true, and we have no great quarrel with Mr. Schumer's modest proposal, nor fear of the future debates to which it might lead, with one proviso. Congress needs, before it votes, to understand what it is doing. The benefit Mr. Schumer proposes is small, but the implications are large and the decision to cross the divide should be witting.