Neo4j 2.0.1 Cypher performance difference between using start and match with a predicate - indexing

Question

Started using Cypher about a week ago (really like it). In the 'browser' interface I'm running two queries:
1) start n=node:Node(name="foo") match (n)-[r*..4]-(m) return n,m
2) match (n{name:"foo"})-[r*..4]-(m) return n,m
The first query returns almost immediately, the second query more than an hour and counting. Naively I would think these would be equivalent, clearly they are not. I ran a 'smaller' (path just up to 1) version of both in the neo-shell so I could profile them.
profile start n=node:Node(name="foo") match (n)-[r*..1]-(m) return n,m;
ColumnFilter(symKeys=["m", "n", " UNNAMED51", "r"], returnItemNames=["n", "m"], _rows=4, _db_hits=0)
TraversalMatcher(start={"expr": "Literal(foo)", "identifiers": ["n"], "key": "Literal(name)",
"idxName": "Node", "producer": "NodeByIndex"}, trail="(n)-[*1..1]-(m)", _rows=4, _db_hits=5)
.
profile match (n{name:"foo"})-[r*..1]-(m) return n,m
ColumnFilter(symKeys=["n", "m", " UNNAMED33", "r"], returnItemNames=["n", "m"], _rows=4, _db_hits=0)
Filter(pred="Property(n,name(0)) == Literal(foo)", _rows=4, _db_hits=196870)
TraversalMatcher(start={"producer": "AllNodes", "identifiers": ["m"]},
trail="(m)-[*1..1]-(n)", _rows=196870, _db_hits=396980)
From other stackoverflow questions I understand db_hits is good to look at, so it looks like the second query has basically done a linear scan (my db is almost 400k nodes). This seems to be confirmed by the "producer" value of "AllNodes" instead of "NodeByIndex".
Obviously I need to specify the match (predicate) differently so that it hits the index. The index is called 'Node' on parameter 'name'. My googling, stacko search is failing me.. how do I specify the conditional in the match so that it hits the index?
Update:
After some poking around it appears I'm using a 'legacy' index? and then trying to hit that with the 'new style (don't use start) query... (kinda extrapolating here). So I can do the following:
create index ON :label(name)
and that would provide an index for a particular label on the name property, but I really want an index (I guess non-legacy index) on ALL the node names. I have use cases where that's important (user may not know the label but does know the name).
Any suggestions or guidance is much appreciated.

Answer 1

Right now there is no global schema index, so you would probably want to create an index on a generic label like Entity or Node and create an index like this:
create index on :Entity(name);
And add that Entity label to all your nodes.
match (n) set n:Entity;