Since the beginning of the past century, the nature of HD108 has been a sub
ject of intense debate. One after another, astronomers explored its variabi
lity and attributed it either to binarity, or to changes in the stellar win
d of a single star. In this article, we analyse a 30 year campaign of spect
roscopic observations of this star with special emphasis on the last 15 yea
rs during which photographic plates have been replaced by CCD detectors. Ou
r investigation of the radial velocities of HD108 yields no significant sho
rt- or long-term period and does not confirm the published periodicities ei
ther. Though the radial velocity of HD108 appears clearly variable, the var
iations cannot be explained by the orbital motion in a spectroscopic binary
. However, our data reveal spectacular changes in the H I Balmer lines and
some He I profiles over the years. These lines continuously evolved from P
Cygni profiles to "pure" absorption lines. A similar behaviour has already
been observed in the past, suggesting that these changes are recurrent. HD1
08 seems to share several characteristics of Oe stars and we discuss differ
ent hypotheses for the origin of the observed long-term variations. As we a
re now in a transition period, a continuous monitoring of HD108 should be c
onsidered for the next few years.