Im having a few issues wrapping my head around how schemas should be implemented.

If I do a default install of openldap (2.4.23) on CentOS 6 the following schemas are automatically included:

cn={0}corba.ldif cn={1}core.ldif cn={2}cosine.ldif cn={3}duaconf.ldif cn={4}dyngroup.ldif cn={5}inetorgperson.ldif cn={6}java.ldif cn={7}misc.ldif cn={8}nis.ldif cn={9}openldap.ldif cn={10}ppolicy.ldif cn={11}collective.ldif

I assume this is done because it allows for a fairly flexible directory that Just Works for nearly everyone. However, many of these schemas contain attributes that I do not use, in fact I can cut down the list of schemas that contain attributes I knowingly use to the following list:

cn={0}core.ldif cn={1}cosine.ldif cn={2}inetorgperson.ldif cn={3}nis.ldif cn={4}customschema.ldif

My issue is that I am not sure if there is any benefit for removing the unused schemas (i.e. I could have just added my customschema as cn={12}customschema.ldif to the default install but I was worried there would be a performance penalty or security issue with including the unused schemas).

Is it advisable to remove unused schemas? Does including unused schemas result in any sort of performance penalty (e.g. while booting or when doing searches)? If I want to use syncrepl to replicate my data do I have to have the exact same list of schemas in the same order on the consumer (i.e. does the number in the curly braces matter)? If I add a schema to a consumer does it have to be added to the provider as well even if the attributes it contains are unused?