On 12/06/2016 02:40 PM, Joshua D. Drake wrote:
> On 12/06/2016 11:12 AM, Eric Schwarzenbach wrote:
>> On 12/06/2016 01:34 PM, Joshua D. Drake wrote:
>>> On 12/06/2016 10:30 AM, Rich Shepard wrote:
>>>
>>>> My thinking is to not store these documents in the database, but to
>>>> store
>>>> them in subdirectories outside the database.
>>>>
>>>> Your thoughts?
>>>
>>> Due to the widely variable size of a PDF document, I would say no. I
>>> would store the metadata and file location.
>>>
>>>
>> Can you elaborate on this? Why is the variable size an issue?
>
> Because it will use at least that size in memory to deliver the
> document to you. Consider a 100MB PDF (not at all uncommon), now
> imagine 40 connections requesting that PDF.
>
Are you sure the whole thing necessarily gets pulled into memory? JDBC
and ODBC support streaming on their BLOB interfaces and isn't the whole
point of this that an application can stream large files a chunk at a
time, the same way it would from the file system? Of course if the db
engine always pulls the whole thing into memory to work with it
regardless of the API, that's another thing, but that wouldn't seem like
a very good design, and I have more faith in the PostgreSQL developers
than that...but I'd certainly like to know for sure.
Cheers,
Eric
