SONAR accepts any file format but the format decides two practical things: whether the file will still be readable in twenty years, and whether its content is findable by a search engine today.
Formats recommended for preservation
Open, documented, widely implemented formats age well. Proprietary formats tied to one vendor's software age badly.
| Content | Recommended | Acceptable | Avoid as the only copy |
|---|---|---|---|
| Text, articles, theses | PDF/A (.pdf) |
PDF, Unicode text (.txt), ODF (.odt) |
.doc, .docx, .pages |
| Structured data | .csv, .xml, .json |
.ods |
.xls, .xlsx, .accdb |
| Images | .tiff, .png, .jpg |
.svg |
.psd, proprietary raw formats |
| Sound | .wav, .flac |
.mp3 |
vendor-specific formats |
| Moving images | .mkv, .mp4 |
— | vendor-specific formats |
PDF/A is the archival profile of PDF: it embeds its fonts, forbids external dependencies, and is the safe default for a publication. Most word processors can export to it directly.
What makes a file searchable
SONAR indexes the text content of the files it can read, which is what allows a reader to find a document by a sentence buried in its body rather than by its title alone.
This extraction works on PDF files only, and only when the PDF carries a text layer:
- a PDF exported from a word processor carries one, and is fully searchable;
- a PDF made by scanning paper does not, unless it has been passed through OCR. It is stored and served normally, but it is invisible to full-text search;
- any other format (
.docx,.zip, images, audio) is stored and served normally, but its content is not indexed.
For example, a scanned thesis with no OCR layer is only findable through the metadata a librarian typed. If a document is not coming up on a phrase you know it contains, this is almost always why. Re-uploading it changes nothing: the file has to be OCRed first.
Size limits
The deposit form refuses a file larger than 500 MiB, and a document cannot carry more than 500 files. Depending on how your instance is deployed, a lower limit may apply. If a large upload fails without a clear message, ask your administrator.
For a very large dataset, deposit a descriptive record in SONAR and point to a data repository designed for it, rather than uploading the whole set.
Several files on one document
A document can carry several files: the publication itself, its supplementary material, a table of contents. Give each one a clear Label, and move the main file to the top of the list with the arrows beside it (the first file is the one used for the thumbnail and presented first to the reader). See File management.