Problem

When uploading certain PowerPoint (.pptx) files to Jira or Confluence Cloud, the system incorrectly identifies the file as a generic application/zip (archive) instead of a PowerPoint presentation. This causes the Media Viewer to display the internal XML structure of the file (folders like ppt/, _rels/) instead of rendering the slides.

This occurs because the MIME type matcher prefers a generic ZIP identification when byte-level sampling is inconclusive, ignoring the valid .pptx extension.

Environment

Atlassian Cloud (Jira and Confluence) using the Atlassian Media API.

Steps to Reproduce

1. Upload a large PowerPoint (.pptx) file to a Jira issue or Confluence page.
2. Once uploaded, click on the attachment to open the Media Viewer preview.
3. Observe the file icon in the attachment list and the content displayed in the previewer.

Expected Results

The file is recognized as a PowerPoint document. The previewer renders the slides, and the file displays the standard PowerPoint icon.

Actual Results

The file is identified as "archive" (application/zip). The previewer shows a list of XML files and folders (the internal structure of the OOXML file), and the icon appears as a generic ZIP/archive folder.

Workaround

• **Re-save the file:** Opening the file in a local PowerPoint application and using "Save As" to create a new version can sometimes reorder the internal metadata, allowing for correct detection upon re-upload.
• *Third-party conversion:* Uploading the file to other systems, like Google Drive and re-downloading it as a PPTX before uploading to Atlassian Cloud has also been seen to "fix" the metadata structure, but not guaranteed.

Notes

Root cause analysis indicates a gap in the MIME type matching logic. When byte-level detectors return application/octet-stream and stream-level detectors return application/zip, the system currently defaults to ZIP even if the filename extension indicates an OOXML type (application/vnd.openxmlformats-officedocument.presentationml.presentation).