Unversioned Ensembl aliases are not unique, except when included with the namespace.
Having Ensembl versioned namespaces is expensive, apx. 400k aliases for each release. With 20 releases, that's a 20x expansion in Ensembl alias size.
Versions have been available since e83. It's time to drop support for unversioned aliases and, therefore, versioned Ensembl releases. SeqRepo will now use the Ensembl namespace (rather than a versioned Ensembl-## namespace).
[x] Remove Ensembl release <= 84. This sidesteps issues regarding unversioned accessions and a problem with ambiguous identifiers for e84 protein accessions.
[x] Remove Ensembl aliases that start with GENSCAN, KI, or GL
[x] Ensure that no unversioned Ensembl aliases remain
[x] Collapse remaining Ensembl-nn aliases into a single Ensembl namespace, preserving history
[x] Require that Ensembl aliases are versioned on loading
Unversioned Ensembl aliases are not unique, except when included with the namespace. Having Ensembl versioned namespaces is expensive, apx. 400k aliases for each release. With 20 releases, that's a 20x expansion in Ensembl alias size.
Versions have been available since e83. It's time to drop support for unversioned aliases and, therefore, versioned Ensembl releases. SeqRepo will now use the
Ensembl
namespace (rather than a versionedEnsembl-##
namespace).