Following the development of high-throughput sequencers, environmental prokaryotic communities could be described by metabarcoding with genetic markers on the 16s domain. However, usual short-read sequencing encounters a limitation in phylogenetic coverage and taxonomic resolution, due to the primers choice and read length. On these critical points, nanopore sequencing, a rising technology, suitable for long-read metabarcoding, was much undervalued because of its relatively higher error rate per read. Here we compared the prokaryotic community structure in samples obtained by short-read metabarcoding on 16sV4-V5 marker (ca. 0.4kbp) analyzed by sequencing-by-synthesis (Illumina dye sequencing, MiSeq), with those obtained by nanopore long-read metabarcoding on bacterial nearly complete 16s (ca. 1.5 kbp, Oxford Nanopore, MinION, R9.2), i.e. a mock community and 52 sediment samples from two contrasted mangrove sites. Nanopore and Illumina retrieved all the bacterial genus from the mock, although both showing similar deviations from the awaited proportions. From the sediment samples, with a coverage-based rarefaction of reads and after singleton filtering, Illumina and Nanopore recorded 34.7% and 35.4% of unknown OTUs, respectively. Nanopore detected 92.2% of the 309 families detected by Illumina, 87.7% of the 448 genus, and recorded 973 additional taxa not detected by Illumina, among which 91.7% were identified to the genus rank. In spite of primer specificities and read length, probably accountable for these discrepancies, co-inertia and Procrustean tests showed that community structures were significantly similar between technologies, showing both a marked contrast between sites and a coherent sea-land orientation within sites.